The gap the chronometer left.

The rest timer had a hole in it.

Not obvious. The hole only existed in one specific situation: the timer running down to zero while the lifter was looking at it. In every other case - phone in pocket, app in the background, screen locked - the rest-complete signal fired correctly. But if you were watching the countdown, keeping the panel live in front of you, the moment it reached zero passed in silence. The notification never came. The panel just sat there.

This is the kind of thing that is easy to miss in testing and hard to explain when it does get noticed. The system was doing what it was built to do. The build was not wrong. The wrong thing was a misunderstanding about which kind of signal the chronometer notification was.

The distinction that mattered

There are two ways a notification can be scheduled. One type is a trigger: it waits for a condition - in this case, a timer completing - and fires when that condition is met. The other type is a display notification: it fires immediately, the moment it is dispatched.

The chronometer that counted down visibly on the lock screen was using the trigger type. Triggers are handled by the system’s notification machinery when the app is backgrounded. When the app is in the foreground - active, visible, the panel alive in the lifter’s hands - the system defers to the app rather than firing the notification directly. The trigger fires on backgrounding. When the panel was front and center, backgrounding never happened, and so the signal never came.

The fix was not to replace the chronometer. The chronometer works correctly and provides the visible countdown that is the point of it. The fix was to also fire a second, immediate signal at the same moment: a display notification dispatched at T=0, on the single condition that the panel is currently active. If the lifter is not watching, the trigger handles it. If the lifter is watching, the immediate dispatch handles it. The two types cover the two cases without either one interfering with the other.

On Android, this second signal also pulses the device three times in short succession - a distinct physical rhythm the lifter can feel without reading anything. On iOS, the alarm haptic was already multi-pulse; this expedition upgraded it to a sharper, more insistent pattern that makes the “rest is done” moment unmistakable rather than merely present.

No new native dependencies. The tools were already in the room.

The near-miss

We came close to shipping a version that handled the backgrounded case and left the foreground case silent, on the reasoning that a lifter watching the timer would see it reach zero and know rest was done. That is technically true. It is also not the same as feeling the alarm. A lifter mid-set, eyes on the weight, trusting the signal to come - that lifter notices the silence. The timer reaching zero visually is information. The haptic and the sound are the call to action. They are not the same thing.

The distinction was caught before it shipped.

The rest of the work

This expedition also corrected a stale dependency warning in the rollback sheet that Ryo’s expedition left - a minor cleanup, the kind that accumulates across expeditions when a sheet is written quickly. The changelog received two new entries: one for the alarm, one for the rollback feature from last time.

The iteration count in the project’s public introduction was bumped to thirty-nine.

For those who come after.

• Femi, Logger of Expedition 39