Close the live-mirror flow hand-off gaps that skipped a timed phase

Field incident (2026-07-17): a Cardio flow run driven from a locked phone
durably completed the 30-min Main Circuit ~5 minutes in, then fought the
watch until manual swipes forced agreement. Second bulletproofing pass
(BULLETPROOFING.md, "2026-07-18 — Live-run mirror & flow hand-off"):

- H4: RunFlowView (both platforms) follows the driver's cross-log
  hand-off frame (forward-only by plan order, runnable target only), and
  a remote completion mid-flow hands off via onAdvance instead of
  dismissing the run (which tore down the Live Activity and stranded the
  flow in the mirror cover).
- H5: wall-clock plausibility gate — a duration exercise never
  auto-completes before startedAt + sets × duration; the flow-terminal
  rest self-repairs to computeResume() and the Finish auto-Done holds
  for a human tap. Human Done taps stay ungated.
- M6: both bridges persist liveVersion as a UserDefaults high-water mark
  so an app relaunch can't get every subsequent frame dropped as stale.
- M7: phone list + mirror cover absorbs gain the watch's strictly-newer
  guard, so a stale cache echo can't regress the working doc.
- Diagnostics: run-flow os_log breadcrumbs (page transitions, applied
  frames, repairs, hand-offs, gate trips) on both drivers.

Open, documented: M5 (no reconciliation signal across a single-set
duration phase), M8 (watch context apply lacks per-log staleness guard),
L5 (spurious pager write-backs read as human swipes).

BULLETPROOFING.md

@@ -1,6 +1,6 @@
 # Bulletproofing the Watch/Phone UX — Findings
 
-An adversarial gap analysis of the iPhone ↔ Apple Watch communication layer, aimed at making the cross-device workout UX bulletproof. Architecture reference: `DEVICE-COMMUNICATION.md` (transports T1–T4, exchanges P1–P9 / W1–W4). Analysis date: 2026-07-09.
+An adversarial gap analysis of the iPhone ↔ Apple Watch communication layer, aimed at making the cross-device workout UX bulletproof. Architecture reference: `DEVICE-COMMUNICATION.md` (transports T1–T4, exchanges P1–P9 / W1–W4). Analysis date: 2026-07-09. **Second pass 2026-07-18** (live-run mirror & flow hand-off — see the dated section at the end), prompted by a field incident.
 
 Findings are ranked by user impact. Each carries code evidence, the concrete failure scenario, and a fix direction. Statuses: **open** / **fixed** (update in place as items land).
 
@@ -144,6 +144,67 @@ Observability: Console.app → iPhone → subsystem `dev.rzen.indie.Workouts` sh
 
 ---
 
+# 2026-07-18 — Live-run mirror & flow hand-off
+
+Second-pass analysis prompted by a field incident (2026-07-17, Cardio flow routine: Warm Up → Main Circuit → Cool Off, all single-set duration exercises): run started on the phone, phone locked in pocket, watch ran the flow correctly — but ~5 min into the 30-min Main Circuit the phone was already counting down Main Circuit's terminal "Coming up: Cool Off" rest, then durably completed the phase early; both devices fought (dismiss ↔ resurrect) until manual swipes forced agreement.
+
+Design context: both devices run the *same* local phase engine; only **human** transitions are broadcast as `LiveProgress` frames — auto-advances are reached independently off shared wall-clock anchors. Flow routines cross exercises via `RunFlowView.advance`, and the incoming exercise broadcasts one hand-off frame.
+
+### H4 — Remote completion dismisses the flow instead of handing off; cross-log frames are ignored  ·  **fixed**
+
+> **Fix (2026-07-18):** `RunFlowView` (both platforms) now follows a frame for a *different* log of the same flow workout by advancing to it — forward-only by plan order, and only into a still-runnable log, so a late-delivered frame for an earlier exercise can't yank the flow backwards. And `ExerciseProgressView`'s remote-resolve observer now hands off (`onAdvance(nextUnfinished)`) instead of dismissing when the current exercise completes remotely mid-flow; dismissal remains for skipped/last-exercise/non-flow cases.
+
+Two halves of one gap. (a) `RunFlowView` filtered incoming frames to `currentLogID` only (phone `RunFlowView.swift:67`, watch `:54`) — the driver's hand-off frame for the *next* exercise was silently dropped, so a follower could only cross an exercise boundary via its own timers (which on the phone need background runtime). (b) When the peer's durable "exercise completed" write landed first, `.onChange(of: log?.status)` called `dismiss()` (phone `ExerciseProgressView.swift:520`, watch `:352`) — popping the whole run, tearing down the Live Activity, sending `liveEnded`, and re-entering later through the mirror cover (a host with different terminal behavior — M3's residual). Which path fired was a pure race between the local timer tick and the ingest of the peer's write inside the same background wake.
+
+### H5 — The between-exercise rest completes a phase it never ran  ·  **fixed**
+
+> **Fix (2026-07-18):** wall-clock plausibility gate on *automatic* completion of duration exercises (both platforms): an auto-fire earlier than `startedAt + sets × duration − slack` is provably wrong — the flow-terminal rest then self-repairs by jumping back to `computeResume()` instead of completing, and the Finish page's auto-Done holds at zero awaiting a human tap. Human Done taps are never gated.
+
+The terminal "Coming up" rest fired `completeExercise()` purely because its countdown window elapsed — regardless of how the page was reached. For a **1-set** exercise any single forward page slip lands on the terminal page (work + 1), so one spurious transition (see L5) durably completed a 30-minute phase ~45 s later; the write then propagated to the watch, whose driver saw `completed` and popped, while its own re-records resurrected the log through the merge — the observed tug-of-war. Likely epicenter of the incident.
+
+### M5 — Reconciliation dead zone across a single-set duration phase  ·  open (mitigated by H5)
+
+Between "started" and "completed" of a 1-set duration exercise there is **no durable write** (`recordProgress` caps at `setCount − 1 = 0`) and **no frame** (all transitions are auto) — for a 30-min phase, a half-hour window in which nothing corrects a desynced peer. `repairFromDurable` can't help: it only jumps set boundaries, and a 1-set exercise has none. A lost hand-off frame (`sendMessage` is reachable-only; the backoff retry gives up after 5 attempts ≈ 15 s) leaves a late-waking follower anchored at local `now` — minutes of drift with no repair until the next human transition.
+
+**Fix direction:** stage auto-advance frames too (the depth-1 latest-wins slot makes this one message per phase), and/or extend the durable repair to recompute the expected phase from `startedAt + duration` for duration exercises. H5's gate already bounds the worst outcome (early durable completion).
+
+### M6 — `liveVersion` resets on app restart; the peer then drops every frame  ·  **fixed**
+
+> **Fix (2026-07-18):** both bridges persist the version as a high-water mark in `UserDefaults` (stamped on send, raised on receive), so a relaunched app keeps outranking what the peer has already seen.
+
+The frame sequence lived in the bridge instance (`PhoneConnectivityBridge.swift:39`, `WatchConnectivityBridge.swift:44`). If the sender's app was killed mid-run (watchdog, memory), its next frames restarted at v1 while the receiver's `lastAppliedVersion` for the run was higher — `isNewer` rejected every one, and since a pocketed phone makes no human transitions to bump the sender's counter back up, the mirror stayed dead for the rest of the run, silently.
+
+### M7 — Phone-side absorbs lack the watch's strictly-newer guard  ·  **fixed**
+
+> **Fix (2026-07-18):** the phone list absorb and the mirror cover absorb now require `incoming > doc.updatedAt`, matching the watch (`Workouts Watch App/Views/WorkoutLogListView.swift:115` and its rationale comment).
+
+`WorkoutLogListView.swift:174` (phone) and `LiveRunCoverView.swift:92` re-seeded the working doc on *any* `updatedAt` change, including a regression (stale cache echo, late-delivered older file version). The next local save then re-stamped the stale content as newest — exactly the clobber the watch-side guard exists to prevent, but on the device that owns the durable write path.
+
+### M8 — Context push can roll the watch cache back per-log  ·  open
+
+`WatchCacheApplier` → `CacheMapper.upsertWorkout` applies the phone's copy with **no per-log staleness check** — the watch-side counterpart of H1's per-log merge exists only on the phone ingest path (`SyncEngine.ingestFromWatch`). The open doc is shielded by the strictly-newer absorb, but only at *workout* granularity: a push newer overall yet stale for one log regresses that log in the cache mid-run. Mostly self-healing (the phone's ingest of the watch's next write re-merges and re-pushes), but it can flap the watch UI and feeds L4's rollback mechanics.
+
+**Fix direction:** per-log `modTime` guard in the watch apply (or full merge symmetry with the phone).
+
+### L5 — Spurious pager write-backs are indistinguishable from human swipes  ·  open
+
+Any `TabView` selection write-back not caused by a real swipe takes the `pageChangeCause == nil` branch in the page observer (both platforms) — recorded *and broadcast* as if the user swiped. One instance (the first-layout snap-to-0) is already guarded by `didRestorePage`, but a later relayout snap (foregrounding, page-count change) would broadcast a wrong page to the peer — and on a 1-set exercise any forward slip is the terminal page (see H5). H5's gate plus the new transition logging bound the damage; a structural fix (distinguishing UIKit write-backs from human interaction) is deferred.
+
+### Diagnostics (added 2026-07-18)
+
+Both platforms' `ExerciseProgressView` now log every page transition (old → new, cause, anchors), every applied remote frame (version/phase/set), durable repairs, remote-resolve hand-offs, and H5 gate trips under subsystem `dev.rzen.indie.Workouts`, category **`run-flow`** — the next incident should be attributable from Console.app instead of reconstructed.
+
+### Recommended order (this pass)
+
+1. **H4 + H5** — the incident killers (follow hand-offs; never auto-complete a phase that provably hasn't run). ✅
+2. **M7, M6** — cheap correctness parity + restart survival. ✅
+3. **Diagnostics** — `run-flow` logging on both platforms. ✅
+4. **M5** — auto-advance keep-alive frames + duration-aware repair (design change; do deliberately).
+5. **M8** — per-log guard in the watch context apply.
+6. **L5** — structural cause detection for pager write-backs.
+
+---
+
 ## Recommended order
 
 The common theme of every HIGH: **the watch doesn't own its session lifecycle — it's a passive recipient of a fire-and-forget launch.** Fix in this order:

CHANGELOG.md

@@ -1,5 +1,9 @@
 **July 2026**
 
+Fixed a rare timing glitch that could skip a long timed phase — like a cardio main circuit — when a workout ran on iPhone and Apple Watch together.
+
+When one device finishes an exercise and moves on, the other now follows along to the next exercise instead of dropping out of the run.
+
 Workouts is now on the Mac: a native app for managing routines and schedules and browsing the exercise library, synced through the same iCloud Drive.
 
 Rest time now adjusts in one-second steps — in Settings and in a routine's custom rest — instead of five-second jumps.

Workouts Watch App/Connectivity/WatchConnectivityBridge.swift

@@ -41,7 +41,12 @@ final class WatchConnectivityBridge: NSObject {
     /// Monotonic sequence stamped on each live-run frame we send. Bumped to stay ahead of any
     /// frame we *receive*, so the two devices share one increasing per-run sequence and either
     /// side can drop a stale / out-of-order delivery (see `LiveProgress.version`).
-    private var liveVersion = 0
+    /// Persisted as a high-water mark: an app relaunch resetting it to 0 would stamp frames
+    /// the peer has already outranked — and with no human transition on the peer to catch
+    /// the counter back up, every subsequent frame would be dropped as stale.
+    private var liveVersion = UserDefaults.standard.integer(forKey: liveVersionDefaultsKey) {
+        didSet { UserDefaults.standard.set(liveVersion, forKey: liveVersionDefaultsKey) }
+    }
 
     /// The latest live-run message we haven't confirmed reached the phone (depth 1, latest-wins).
     /// Staged regardless of reachability and re-sent by `flushLive()` when reachability/activation
@@ -451,3 +456,6 @@ private enum PendingLive {
     case progress(LiveProgress)
     case ended(workoutID: String, logID: String)
 }
+
+/// Defaults key for the persisted live-frame version high-water mark (see `liveVersion`).
+private let liveVersionDefaultsKey = "liveMirrorVersionHighWater"

Workouts Watch App/Views/ExerciseProgressView.swift

@@ -5,6 +5,7 @@
 // Copyright 2025 Rouslan Zenetl. All Rights Reserved.
 //
 
+import os
 import SwiftUI
 import WatchKit
 
@@ -23,6 +24,10 @@ import WatchKit
 /// A dot row tracks progress with one marker per work set — the active set drawn as a
 /// wider dash, with the gap widening at the rest you're currently in.
 struct ExerciseProgressView: View {
+    /// Run-flow breadcrumbs (page transitions, applied frames, repairs) — the mirror's
+    /// failure modes are cross-device timing races, unattributable without a trail.
+    private static let log = Logger(subsystem: "dev.rzen.indie.Workouts", category: "run-flow")
+
     @Environment(\.dismiss) private var dismiss
 
     /// Live HR straight off the session's builder — drives the target-HR readout when
@@ -321,6 +326,7 @@ struct ExerciseProgressView: View {
             guard didRestorePage else { return }
             let cause = pageChangeCause
             pageChangeCause = nil
+            Self.log.info("page \(oldPage) → \(newPage) cause=\(cause.map { "\($0)" } ?? "human", privacy: .public) log=\(logID, privacy: .public)")
             switch cause {
             case .remote:
                 // The phone already recorded and owns this transition — just follow it.
@@ -351,13 +357,23 @@ struct ExerciseProgressView: View {
         }
         .onChange(of: log?.status) { _, raw in
             // The log resolved out from under us (the run was completed or ended early on
-            // the phone). Leave the live flow — staying would let the next recorded set
-            // write a resolved log back to in-progress, resurrecting it through the merge.
-            // `locallyResolved` keeps our own Done / flow hand-off from re-dismissing.
+            // the phone). Staying would let the next recorded set write a resolved log
+            // back to in-progress, resurrecting it through the merge. In flow mode a
+            // remote *completion* means the driver has already handed off — follow it to
+            // the next unfinished exercise instead of falling out of the run. Everything
+            // else (skipped, last exercise, non-flow) leaves the flow as before.
+            // `locallyResolved` keeps our own Done / flow hand-off from re-triggering this.
             guard didRestorePage, !locallyResolved,
                   let status = raw.flatMap(WorkoutStatus.init(rawValue:)),
                   status == .completed || status == .skipped else { return }
-            dismiss()
+            locallyResolved = true
+            if autoAdvance, status == .completed, let next = nextUnfinishedLogID, let onAdvance {
+                Self.log.info("remote completion of \(logID, privacy: .public) → flow hand-off to \(next, privacy: .public)")
+                onAdvance(next)
+            } else {
+                Self.log.info("remote resolve of \(logID, privacy: .public) (\(status.rawValue, privacy: .public)) → dismiss")
+                dismiss()
+            }
         }
         .onAppear {
             guard !didRestorePage else { return }
@@ -444,6 +460,7 @@ struct ExerciseProgressView: View {
         lastAppliedStart = frame.phaseStart
         if frame.setCount != setCount { setCount = frame.setCount }
         let target = page(forPhase: frame.phase, setIndex: frame.setIndex)
+        Self.log.info("apply frame v\(frame.version) \(frame.phase.rawValue, privacy: .public)/\(frame.setIndex) → page \(target) log=\(logID, privacy: .public)")
         // A followed frame is progress this screen now *knows*, so a durable echo of the
         // same transition can't re-trigger the repair jump (a remote reset re-baselines).
         knownStateIndex = frame.phase == .ready ? 0 : max(knownStateIndex, completedSets(for: target))
@@ -471,6 +488,7 @@ struct ExerciseProgressView: View {
         if log.sets > setCount { setCount = log.sets }
         let target = base + min(max(0, completed), setCount - 1) * 2
         guard target > currentPage else { return }
+        Self.log.info("durable repair: completed=\(completed) → page \(target) log=\(logID, privacy: .public)")
         clearAnchor()
         pageChangeCause = .remote
         withAnimation { currentPage = target }
@@ -482,6 +500,33 @@ struct ExerciseProgressView: View {
         pageAnchor = nil
     }
 
+    /// Wall-clock plausibility gate for *automatic* completion of a duration exercise: its
+    /// work windows alone take `sets × duration`, so an auto-fire earlier than that (minus
+    /// slack) proves the terminal page was reached erroneously — a spurious pager
+    /// write-back or a stale anchor, not a run that actually happened. Human Done taps are
+    /// never gated; rep-based exercises have no wall-clock expectation and always pass, as
+    /// does a log missing its `startedAt` stamp (nothing to judge against).
+    private var autoCompletionPlausible: Bool {
+        guard isDuration, let log, let started = log.startedAt else { return true }
+        let minimum = Double(workDurationSeconds * setCount) - 2
+        return Date().timeIntervalSince(started) >= minimum
+    }
+
+    /// The flow-terminal rest fired before the exercise could have run (see
+    /// `autoCompletionPlausible`) — jump back to where the durable timestamps say the run
+    /// actually is instead of durably completing a phase that never happened.
+    private func repairEarlyTerminal() {
+        let resume = computeResume()
+        Self.log.warning("early terminal auto-fire blocked for \(logID, privacy: .public); repairing to page \(resume.page)")
+        guard currentPage != resume.page else {
+            pageAnchor = resume.anchor
+            return
+        }
+        pageAnchor = resume.anchor
+        pageChangeCause = .remote
+        withAnimation { currentPage = resume.page }
+    }
+
     /// Inverse of `liveSnapshot`'s page→frame mapping: a frame's phase/set → page index.
     private func page(forPhase phase: LiveRunPhase, setIndex: Int) -> Int {
         let set = min(max(0, setIndex), max(0, setCount - 1))
@@ -611,6 +656,9 @@ struct ExerciseProgressView: View {
                         anchorStart: anchorStart,
                         anchorEnd: anchorEnd
                     ) { _ in
+                        // An auto-fire before the exercise could possibly have run proves
+                        // this page was reached erroneously — repair, don't complete.
+                        guard autoCompletionPlausible else { repairEarlyTerminal(); return }
                         completeExercise()
                         // Re-resolve at fire time — the next exercise may have been
                         // finished elsewhere while we rested. None left ⇒ end the run.
@@ -629,7 +677,8 @@ struct ExerciseProgressView: View {
                         isActive: isActive,
                         onDone: { completeExercise(); dismiss() },
                         onOneMore: addSet,
-                        anchorEnd: anchorEnd
+                        anchorEnd: anchorEnd,
+                        autoFireAllowed: { autoCompletionPlausible }
                     )
                 }
             } else if cycleIndex.isMultiple(of: 2) {
@@ -1191,6 +1240,11 @@ private struct FinishPhaseView: View {
     /// sender's wall clock instead of local `now + countdown`, so both devices fire together.
     var anchorEnd: Date? = nil
 
+    /// Gates the *automatic* Done only (the tap always works): when false, the countdown
+    /// holds at zero awaiting a human — the host's wall-clock plausibility check says the
+    /// exercise can't genuinely be finished yet (see `autoCompletionPlausible`).
+    var autoFireAllowed: () -> Bool = { true }
+
     @AppStorage("doneCountdownSeconds") private var doneCountdownSeconds: Int = 5
 
     /// Wall-clock deadline for the auto-Done (same Always-On rationale as the rest timer).
@@ -1236,7 +1290,7 @@ private struct FinishPhaseView: View {
         guard isActive, !didFire else { return }
         let r = Int(ceil(endDate.timeIntervalSinceNow))
         remaining = max(0, r)
-        if r <= 0 { fire() }
+        if r <= 0, autoFireAllowed() { fire() }
     }
 
     private func fire() {

Workouts Watch App/Views/RunFlowView.swift

@@ -59,6 +59,26 @@ struct RunFlowView: View {
         }
         .onAppear { bridge.navigatedRunID = currentLogID }
         .onDisappear { if bridge.navigatedRunID == currentLogID { bridge.navigatedRunID = nil } }
+        .onChange(of: bridge.liveIncoming) { _, frame in
+            followRemoteHandoff(frame)
+        }
+    }
+
+    /// Follow the driver across an exercise boundary. Auto-advances are never broadcast, so
+    /// a frame for a *different* log of this flow workout is always the peer's discrete
+    /// hand-off into the next exercise — which the per-exercise child can't follow (its
+    /// incoming-frame filter is scoped to `currentLogID`). Forward-only by plan order and
+    /// only into a still-runnable log, so a late-delivered frame for an earlier exercise
+    /// can never yank the flow backwards.
+    private func followRemoteHandoff(_ frame: LiveProgress?) {
+        guard doc.autoAdvance == true,
+              let frame, frame.workoutID == doc.id, frame.logID != currentLogID else { return }
+        let sorted = doc.logs.sorted { $0.order != $1.order ? $0.order < $1.order : $0.id < $1.id }
+        guard let from = sorted.firstIndex(where: { $0.id == currentLogID }),
+              let to = sorted.firstIndex(where: { $0.id == frame.logID }), to > from else { return }
+        let status = WorkoutStatus(rawValue: sorted[to].status) ?? .notStarted
+        guard status == .notStarted || status == .inProgress else { return }
+        advance(to: frame.logID)
     }
 
     /// Hand off to the next exercise. Suppresses the leaving exercise's live-ended teardown

Workouts/Connectivity/PhoneConnectivityBridge.swift

@@ -36,7 +36,12 @@ final class PhoneConnectivityBridge: NSObject {
     /// Monotonic sequence stamped on each live-run frame we send. Bumped to stay ahead of
     /// any frame we *receive*, so the two devices share one increasing sequence per run and
     /// either side can drop a stale / out-of-order delivery (see `LiveProgress.version`).
-    private var liveVersion = 0
+    /// Persisted as a high-water mark: an app relaunch resetting it to 0 would stamp frames
+    /// the peer has already outranked — and with no human transition on the peer to catch
+    /// the counter back up, every subsequent frame would be dropped as stale.
+    private var liveVersion = UserDefaults.standard.integer(forKey: liveVersionDefaultsKey) {
+        didSet { UserDefaults.standard.set(liveVersion, forKey: liveVersionDefaultsKey) }
+    }
 
     /// The latest live-run message we haven't confirmed reached the watch (depth 1, latest-wins).
     /// Staged regardless of reachability and re-sent by `flushLive()` when reachability/activation
@@ -332,3 +337,6 @@ private enum PendingLive {
     case progress(LiveProgress)
     case ended(workoutID: String, logID: String)
 }
+
+/// Defaults key for the persisted live-frame version high-water mark (see `liveVersion`).
+private let liveVersionDefaultsKey = "liveMirrorVersionHighWater"

Workouts/Views/WorkoutLogs/ExerciseProgressView.swift

@@ -5,6 +5,7 @@
 // Copyright 2025 Rouslan Zenetl. All Rights Reserved.
 //
 
+import os
 import SwiftUI
 import UIKit
 
@@ -29,6 +30,10 @@ import UIKit
 /// workout stopwatch in big glanceable digits. A row of phase dots tracks progress:
 /// purple for work, teal for rest, with the current phase drawn as a wider dash.
 struct ExerciseProgressView: View {
+    /// Run-flow breadcrumbs (page transitions, applied frames, repairs) — the mirror's
+    /// failure modes are cross-device timing races, unattributable without a trail.
+    private static let log = Logger(subsystem: "dev.rzen.indie.Workouts", category: "run-flow")
+
     @Environment(\.dismiss) private var dismiss
     @Environment(\.verticalSizeClass) private var verticalSizeClass
     @Environment(\.modelContext) private var modelContext
@@ -486,6 +491,7 @@ struct ExerciseProgressView: View {
             guard didRestorePage else { return }
             let cause = pageChangeCause
             pageChangeCause = nil
+            Self.log.info("page \(oldPage) → \(newPage) cause=\(cause.map { "\($0)" } ?? "human", privacy: .public) log=\(logID, privacy: .public)")
             switch cause {
             case .remote:
                 // The watch already recorded and owns this transition — just follow it.
@@ -519,13 +525,24 @@ struct ExerciseProgressView: View {
         }
         .onChange(of: log?.status) { _, raw in
             // The log resolved out from under us (the run was completed or ended early on
-            // another device). Leave the live flow — staying would let the next recorded
-            // set write a resolved log back to in-progress, resurrecting it through the
-            // merge. `locallyResolved` keeps our own Done / flow hand-off from re-dismissing.
+            // another device). Staying would let the next recorded set write a resolved log
+            // back to in-progress, resurrecting it through the merge. In flow mode a remote
+            // *completion* means the driver has already handed off — follow it to the next
+            // unfinished exercise instead of falling out of the run (which would tear down
+            // the Live Activity and strand the flow in the mirror cover). Everything else
+            // (skipped, last exercise, non-flow) leaves the flow as before.
+            // `locallyResolved` keeps our own Done / flow hand-off from re-triggering this.
             guard didRestorePage, !locallyResolved,
                   let status = raw.flatMap(WorkoutStatus.init(rawValue:)),
                   status == .completed || status == .skipped else { return }
-            dismiss()
+            locallyResolved = true
+            if autoAdvance, status == .completed, let next = nextUnfinishedLogID, let onAdvance {
+                Self.log.info("remote completion of \(logID, privacy: .public) → flow hand-off to \(next, privacy: .public)")
+                onAdvance(next)
+            } else {
+                Self.log.info("remote resolve of \(logID, privacy: .public) (\(status.rawValue, privacy: .public)) → dismiss")
+                dismiss()
+            }
         }
         .onAppear {
             guard !didRestorePage else { return }
@@ -666,6 +683,7 @@ struct ExerciseProgressView: View {
         lastAppliedStart = frame.phaseStart
         if frame.setCount != setCount { setCount = frame.setCount }
         let target = page(forPhase: frame.phase, setIndex: frame.setIndex)
+        Self.log.info("apply frame v\(frame.version) \(frame.phase.rawValue, privacy: .public)/\(frame.setIndex) → page \(target) log=\(logID, privacy: .public)")
         // A followed frame is progress this screen now *knows*, so a durable echo of the
         // same transition can't re-trigger the repair jump (a remote reset re-baselines).
         knownStateIndex = frame.phase == .ready ? 0 : max(knownStateIndex, completedSets(for: target))
@@ -693,6 +711,7 @@ struct ExerciseProgressView: View {
         if log.sets > setCount { setCount = log.sets }
         let target = base + min(max(0, completed), setCount - 1) * 2
         guard target > currentPage else { return }
+        Self.log.info("durable repair: completed=\(completed) → page \(target) log=\(logID, privacy: .public)")
         clearAnchor()
         pageChangeCause = .remote
         withAnimation { currentPage = target }
@@ -704,6 +723,33 @@ struct ExerciseProgressView: View {
         pageAnchor = nil
     }
 
+    /// Wall-clock plausibility gate for *automatic* completion of a duration exercise: its
+    /// work windows alone take `sets × duration`, so an auto-fire earlier than that (minus
+    /// slack) proves the terminal page was reached erroneously — a spurious pager
+    /// write-back or a stale anchor, not a run that actually happened. Human Done taps are
+    /// never gated; rep-based exercises have no wall-clock expectation and always pass, as
+    /// does a log missing its `startedAt` stamp (nothing to judge against).
+    private var autoCompletionPlausible: Bool {
+        guard isDuration, let log, let started = log.startedAt else { return true }
+        let minimum = Double(workDurationSeconds * setCount) - 2
+        return Date().timeIntervalSince(started) >= minimum
+    }
+
+    /// The flow-terminal rest fired before the exercise could have run (see
+    /// `autoCompletionPlausible`) — jump back to where the durable timestamps say the run
+    /// actually is instead of durably completing a phase that never happened.
+    private func repairEarlyTerminal() {
+        let resume = computeResume()
+        Self.log.warning("early terminal auto-fire blocked for \(logID, privacy: .public); repairing to page \(resume.page)")
+        guard currentPage != resume.page else {
+            pageAnchor = resume.anchor
+            return
+        }
+        pageAnchor = resume.anchor
+        pageChangeCause = .remote
+        withAnimation { currentPage = resume.page }
+    }
+
     /// Inverse of `liveSnapshot`'s page→frame mapping: a frame's phase/set → page index.
     private func page(forPhase phase: LiveRunPhase, setIndex: Int) -> Int {
         let set = min(max(0, setIndex), max(0, setCount - 1))
@@ -786,6 +832,9 @@ struct ExerciseProgressView: View {
                         onActivate: announceComingUp,
                         onCountdownBeat: announceCountdownBeat
                     ) { _ in
+                        // An auto-fire before the exercise could possibly have run proves
+                        // this page was reached erroneously — repair, don't complete.
+                        guard autoCompletionPlausible else { repairEarlyTerminal(); return }
                         completeExercise()
                         // Re-resolve at fire time — the next exercise may have been
                         // finished elsewhere while we rested. None left ⇒ end the run.
@@ -805,7 +854,8 @@ struct ExerciseProgressView: View {
                         isActive: isActive,
                         onDone: { completeExercise(); dismiss() },
                         onOneMore: addSet,
-                        anchorEnd: anchorEnd
+                        anchorEnd: anchorEnd,
+                        autoFireAllowed: { autoCompletionPlausible }
                     )
                     .overlay(alignment: .bottom) { adjustPill }
                 }
@@ -1777,6 +1827,11 @@ private struct FinishPhaseView: View {
     /// sender's wall clock instead of local `now + countdown`, so both devices fire together.
     var anchorEnd: Date? = nil
 
+    /// Gates the *automatic* Done only (the tap always works): when false, the countdown
+    /// holds at zero awaiting a human — the host's wall-clock plausibility check says the
+    /// exercise can't genuinely be finished yet (see `autoCompletionPlausible`).
+    var autoFireAllowed: () -> Bool = { true }
+
     @AppStorage("doneCountdownSeconds") private var doneCountdownSeconds: Int = 5
 
     /// Wall-clock deadline for the auto-Done. `remaining` is what the Done button shows.
@@ -1824,7 +1879,7 @@ private struct FinishPhaseView: View {
         if max(0, r) != remaining {
             withAnimation(.snappy(duration: 0.25)) { remaining = max(0, r) }
         }
-        if r <= 0 { fire() }
+        if r <= 0, autoFireAllowed() { fire() }
     }
 
     private func fire() {

Workouts/Views/WorkoutLogs/LiveRunCoverView.swift

@@ -89,10 +89,13 @@ struct LiveRunCoverView: View {
         .onAppear {
             if doc == nil, let workout { doc = WorkoutDocument(from: workout) }
         }
-        .onChange(of: workout?.updatedAt) { _, _ in
+        .onChange(of: workout?.updatedAt) { _, incoming in
             // Absorb the durable progress the watch drove (currentStateIndex / status). The
             // live *page* is driven by frames, not this — so re-seeding the doc is safe.
-            if let workout { doc = WorkoutDocument(from: workout) }
+            // Strictly newer only (matching the list absorbs): a stale cache echo must not
+            // regress the working copy, which the next persist would re-stamp as newest.
+            guard let incoming, incoming > (doc?.updatedAt ?? .distantPast), let workout else { return }
+            doc = WorkoutDocument(from: workout)
         }
     }
 

Workouts/Views/WorkoutLogs/RunFlowView.swift

@@ -79,6 +79,26 @@ struct RunFlowView: View {
             // than in the child's onDisappear — no longer cuts off the next exercise's cue.
             speechAnnouncer?.stop()
         }
+        .onChange(of: liveRun.current) { _, frame in
+            followRemoteHandoff(frame)
+        }
+    }
+
+    /// Follow the driver across an exercise boundary. Auto-advances are never broadcast, so
+    /// a frame for a *different* log of this flow workout is always the peer's discrete
+    /// hand-off into the next exercise — which the per-exercise child can't follow (its
+    /// incoming-frame filter is scoped to `currentLogID`). Forward-only by plan order and
+    /// only into a still-runnable log, so a late-delivered frame for an earlier exercise
+    /// can never yank the flow backwards.
+    private func followRemoteHandoff(_ frame: LiveProgress?) {
+        guard doc.autoAdvance == true,
+              let frame, frame.workoutID == doc.id, frame.logID != currentLogID else { return }
+        let sorted = doc.logs.sorted { $0.order != $1.order ? $0.order < $1.order : $0.id < $1.id }
+        guard let from = sorted.firstIndex(where: { $0.id == currentLogID }),
+              let to = sorted.firstIndex(where: { $0.id == frame.logID }), to > from else { return }
+        let status = WorkoutStatus(rawValue: sorted[to].status) ?? .notStarted
+        guard status == .notStarted || status == .inProgress else { return }
+        advance(to: frame.logID)
     }
 
     /// Hand off to the next exercise. Suppresses the leaving exercise's live-ended/activity

Workouts/Views/WorkoutLogs/WorkoutLogListView.swift

@@ -171,8 +171,13 @@ struct WorkoutLogListView: View {
         // cache reflects them, so the list shows live status on return. The observed
         // expression itself must be liveness-guarded too: it's evaluated on every
         // body pass, and reading a persisted property on a dead @Model traps.
-        .onChange(of: workout.isLive ? workout.updatedAt : nil) { _, _ in
-            if let fresh = WorkoutDocument(fromLive: workout) { doc = fresh }
+        // Strictly-newer guard, matching the watch list: an equal stamp is our own
+        // write echoing back, and a stale cache echo (late-delivered older file
+        // version) must not regress the working copy mid-run — the next local save
+        // would re-stamp that stale content as newest and clobber real progress.
+        .onChange(of: workout.isLive ? workout.updatedAt : nil) { _, incoming in
+            if let incoming, incoming > doc.updatedAt,
+               let fresh = WorkoutDocument(fromLive: workout) { doc = fresh }
         }
         .toolbar {
             ToolbarItem(placement: .primaryAction) {