fix(store): gate ActivitiesMissingWorkout on details already fetched

fillPendingDetails and fillPendingWorkouts are each independently
LIMIT-bounded over different candidate sets, so on a large first
backfill a structured-workout activity could fall inside the workouts
pass's window while still outside the details pass's window.
fillActivityWorkout would then align workout targets against zero laps
(details/laps never written yet) and unconditionally mark
workout_raw_json non-NULL, permanently losing that activity's target
pace/HR bands once its real laps arrived later -- silently, with no
error.

Add details_fetched_at IS NOT NULL to ActivitiesMissingWorkout's and
CountActivitiesMissingWorkout's WHERE clause: an activity is only
eligible for a workout fetch once its laps actually exist to align
against. This still covers the intended retry case (an activity whose
workout fetch previously failed always has details_fetched_at already
set) while excluding never-yet-processed activities.

Rename/rewrite the store test to assert the corrected exclusion
(count=1, not 2) as an explicit regression test, and fix the
TestSyncStatus_ReportsPhaseAwareProgressAndWorkoutsPending API fixture,
which exercised the same buggy shape.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

backend/internal/api/api_test.go

@@ -669,12 +669,20 @@ func TestSyncStatus_ReportsPhaseAwareProgressAndWorkoutsPending(t *testing.T) {
 	ctx := newCtx()
 
 	workoutID := int64(555)
-	if _, err := db.UpsertActivity(ctx, userID, store.Activity{
+	activityID, err := db.UpsertActivity(ctx, userID, store.Activity{
 		GarminActivityID: 1, WorkoutID: &workoutID,
 		StartTimeUTC: "2026-07-11 06:00:00", RawJSON: "{}",
-	}); err != nil {
+	})
+	if err != nil {
 		t.Fatalf("UpsertActivity: %v", err)
 	}
+	// ActivitiesMissingWorkout only surfaces activities whose details have
+	// already been fetched (see store.ActivitiesMissingWorkout) -- an
+	// activity is only eligible for a workout fetch once fillActivityDetails
+	// has given it laps to align target pace/HR bands against.
+	if err := db.SetActivityDetails(ctx, userID, activityID, "{}"); err != nil {
+		t.Fatalf("SetActivityDetails: %v", err)
+	}
 
 	rec := doJSON(t, s.Router(), http.MethodGet, "/api/sync/status", nil)
 	if rec.Code != http.StatusOK {

backend/internal/store/activities.go

@@ -276,16 +276,24 @@ func (db *DB) CountActivitiesMissingDetails(ctx context.Context, userID int64) (
 // ActivitiesMissingWorkout returns userID's activities that have a
 // structured workout (WorkoutID set at initial upsert time, straight from
 // Garmin's activity summary) but haven't had get_workout_by_id fetched yet.
-// Independently queryable from ActivitiesMissingDetails: workout_id is
+// Gated on details_fetched_at IS NOT NULL: fillActivityWorkout resolves each
-// known well before any detail fetch, and workout_raw_json is only ever
+// lap's target pace/HR band via alignWorkoutTargets, which needs the
-// set by SetActivityWorkout, so this also picks up an activity whose
+// activity's laps already written by fillActivityDetails -- without this
-// details were fetched successfully in some prior run but whose workout
+// gate, a structured-workout activity whose details/laps haven't been
-// fetch failed back then -- ActivitiesMissingDetails would never surface
+// fetched yet would fall into a separate, independently LIMIT-bounded batch
-// that activity again (details_fetched_at/splits_fetched_at are already
+// than fillPendingDetails's, get its workout_raw_json set against zero laps
-// set), silently losing its target pace/HR bands forever without this.
+// (a silent no-op alignment), and then never be retried once its laps
+// finally arrive, since workout_raw_json IS NULL is the only signal this
+// query has left. This is still independent of splits_fetched_at (unlike
+// ActivitiesMissingDetails, which requires both): splits/laps aren't needed
+// to resolve workout targets, only details_fetched_at is. It also still
+// covers the intended retry case -- an activity whose workout fetch
+// previously failed after details succeeded always has details_fetched_at
+// already set, so it's still surfaced here -- while excluding activities
+// that simply haven't been processed by fillActivityDetails at all yet.
 func (db *DB) ActivitiesMissingWorkout(ctx context.Context, userID int64, limit int) ([]Activity, error) {
 	rows, err := db.QueryContext(ctx, `SELECT `+activityColumns+` FROM activities
-		WHERE user_id = ? AND workout_id IS NOT NULL AND workout_raw_json IS NULL
+		WHERE user_id = ? AND workout_id IS NOT NULL AND workout_raw_json IS NULL AND details_fetched_at IS NOT NULL
 		ORDER BY start_time_utc DESC LIMIT ?`, userID, limit)
 	if err != nil {
 		return nil, fmt.Errorf("list activities missing workout for user %d: %w", userID, err)
@@ -306,11 +314,12 @@ func (db *DB) ActivitiesMissingWorkout(ctx context.Context, userID int64, limit
 // CountActivitiesMissingWorkout returns how many of userID's activities
 // still need get_workout_by_id fetched, regardless of any per-call batch
 // limit -- used to report overall remaining work, mirroring
-// CountActivitiesMissingDetails.
+// CountActivitiesMissingDetails. See ActivitiesMissingWorkout for why this
+// is additionally gated on details_fetched_at IS NOT NULL.
 func (db *DB) CountActivitiesMissingWorkout(ctx context.Context, userID int64) (int, error) {
 	var n int
 	err := db.QueryRowContext(ctx, `SELECT COUNT(*) FROM activities
-		WHERE user_id = ? AND workout_id IS NOT NULL AND workout_raw_json IS NULL`, userID).Scan(&n)
+		WHERE user_id = ? AND workout_id IS NOT NULL AND workout_raw_json IS NULL AND details_fetched_at IS NOT NULL`, userID).Scan(&n)
 	if err != nil {
 		return 0, fmt.Errorf("count activities missing workout for user %d: %w", userID, err)
 	}

backend/internal/store/store_test.go

@@ -356,7 +356,23 @@ func TestCurrentAssignment_ScopedToOwningUser(t *testing.T) {
 	}
 }
 
-func TestActivitiesMissingWorkout_IndependentOfDetailsFetchStatus(t *testing.T) {
+// TestActivitiesMissingWorkout_OnlyIncludesActivitiesWithDetailsAlreadyFetched
+// is a regression test for a bug where ActivitiesMissingWorkout selected
+// activities purely on workout_id IS NOT NULL AND workout_raw_json IS NULL,
+// with no regard for whether get_activity_details had ever run for that
+// activity. Since fillPendingDetails and fillPendingWorkouts are each
+// independently LIMIT-bounded over different candidate sets, a
+// structured-workout activity could fall inside the workouts pass's window
+// while still outside the details pass's window on a large first backfill.
+// fillActivityWorkout would then call LapsForActivity on an activity with no
+// laps yet written, compute alignWorkoutTargets against zero laps (a
+// no-op), and unconditionally call SetActivityWorkout anyway -- permanently
+// marking workout_raw_json non-NULL before the activity ever had a chance
+// to get real target pace/HR bands once its laps finally arrived. The fix
+// gates the query on details_fetched_at IS NOT NULL, so an activity is only
+// eligible for a workout fetch once fillActivityDetails has actually given
+// it laps to align against.
+func TestActivitiesMissingWorkout_OnlyIncludesActivitiesWithDetailsAlreadyFetched(t *testing.T) {
 	db := openTestDB(t)
 	ctx := context.Background()
 	userID, err := db.ProvisionUser(ctx, "test-sub", "Test")
@@ -365,6 +381,9 @@ func TestActivitiesMissingWorkout_IndependentOfDetailsFetchStatus(t *testing.T)
 	}
 
 	workoutID := int64(999)
+	// Activity a: has a workout_id, but details/splits have never been
+	// fetched (the exact shape of the bug above). Must now be EXCLUDED --
+	// this is the regression assertion for the bug.
 	withWorkoutNoDetails := Activity{
 		GarminActivityID: 1, WorkoutID: &workoutID,
 		StartTimeUTC: "2026-07-01 06:00:00", RawJSON: "{}",
@@ -374,6 +393,11 @@ func TestActivitiesMissingWorkout_IndependentOfDetailsFetchStatus(t *testing.T)
 		t.Fatalf("UpsertActivity (a): %v", err)
 	}
 
+	// Activity b: has a workout_id, and its details/splits have already
+	// been fetched (e.g. its workout fetch failed in some prior run after
+	// details succeeded). This is the one case ActivitiesMissingWorkout must
+	// still surface, since ActivitiesMissingDetails would never pick this
+	// activity up again once details_fetched_at/splits_fetched_at are set.
 	withWorkoutAndDetails := Activity{
 		GarminActivityID: 2, WorkoutID: &workoutID,
 		StartTimeUTC: "2026-07-02 06:00:00", RawJSON: "{}",
@@ -382,10 +406,6 @@ func TestActivitiesMissingWorkout_IndependentOfDetailsFetchStatus(t *testing.T)
 	if err != nil {
 		t.Fatalf("UpsertActivity (b): %v", err)
 	}
-	// Simulate an activity whose details/splits were already fetched in a
-	// prior run, but whose workout fetch failed back then -- this is exactly
-	// the case ActivitiesMissingWorkout must still surface, since it's
-	// queried independently of details_fetched_at/splits_fetched_at.
 	if err := db.SetActivityDetails(ctx, userID, idB, "{}"); err != nil {
 		t.Fatalf("SetActivityDetails (b): %v", err)
 	}
@@ -393,6 +413,7 @@ func TestActivitiesMissingWorkout_IndependentOfDetailsFetchStatus(t *testing.T)
 		t.Fatalf("SetActivitySplitsFetched (b): %v", err)
 	}
 
+	// Activity c: no workout_id at all. Must be excluded.
 	noWorkout := Activity{
 		GarminActivityID: 3, StartTimeUTC: "2026-07-03 06:00:00", RawJSON: "{}",
 	}
@@ -400,6 +421,7 @@ func TestActivitiesMissingWorkout_IndependentOfDetailsFetchStatus(t *testing.T)
 		t.Fatalf("UpsertActivity (c): %v", err)
 	}
 
+	// Activity d: already has its workout_raw_json set. Must be excluded.
 	alreadyHasWorkout := Activity{
 		GarminActivityID: 4, WorkoutID: &workoutID,
 		StartTimeUTC: "2026-07-04 06:00:00", RawJSON: "{}",
@@ -416,23 +438,24 @@ func TestActivitiesMissingWorkout_IndependentOfDetailsFetchStatus(t *testing.T)
 	if err != nil {
 		t.Fatalf("CountActivitiesMissingWorkout: %v", err)
 	}
-	if n != 2 {
+	if n != 1 {
-		t.Fatalf("CountActivitiesMissingWorkout = %d, want 2 (activities a and b)", n)
+		t.Fatalf("CountActivitiesMissingWorkout = %d, want 1 (activity b only -- activity a must be excluded since its details were never fetched)", n)
 	}
 
 	pending, err := db.ActivitiesMissingWorkout(ctx, userID, 10)
 	if err != nil {
 		t.Fatalf("ActivitiesMissingWorkout: %v", err)
 	}
-	if len(pending) != 2 {
+	if len(pending) != 1 {
-		t.Fatalf("ActivitiesMissingWorkout returned %d activities, want 2", len(pending))
+		t.Fatalf("ActivitiesMissingWorkout returned %d activities, want 1", len(pending))
+	}
+	if pending[0].ID != idB {
+		t.Errorf("ActivitiesMissingWorkout = %+v, want only activity b (id %d)", pending, idB)
 	}
-	ids := map[int64]bool{}
 	for _, a := range pending {
-		ids[a.ID] = true
+		if a.ID == idA {
+			t.Errorf("ActivitiesMissingWorkout incorrectly included activity a (workout_id set but details never fetched) -- regression for the silent-loss bug")
 		}
-	if !ids[idA] || !ids[idB] {
-		t.Errorf("ActivitiesMissingWorkout = %+v, want to include activities a and b", pending)
 	}
 }