Long page stalls with no cause in stalls.log, 2–5 November, on this Mac
I matched the bursts against the Mac's power log (pmset -g log) and the login window's log:
| Burst (CET) | Canvas | Stalls | Longest | The screen at that time |
|---|---|---|---|---|
| 2 Nov, 09:07–09:15 | obeya | 17 | 6.1 s | Display off since the evening before, on again at 09:47 |
| 3 Nov, 10:48–10:53 | obeya | 34 | 4.0 s | Display on, but the system counted the user as idle (no input since before 10:40) |
| 3 Nov, 11:55–12:03 | obeya | 29 | 36.0 s | Display off and screen locked at 11:54:10, a minute before the first one |
| 3 Nov, 16:32–16:36 | nemotron | 9 | 9.6 s | Four minutes after the Mac restarted (16:28) |
The rest were single stalls of 0.5–2 s, several of them in the night. While the screen is locked, the lock screen is the app in front (I checked this morning), so no other window has the focus.
A test window with WebKit, in the background, ran the same heartbeat as Obeya's pages. With all 14 cores busy for 30 s, at normal and at low priority, the longest wait for a heartbeat was 2 ms. So a busy machine alone doesn't explain it. What held the real window back while the screen was off stays unknown. A stall in a window nobody can see doesn't matter to the owner either way.
Every page stall in the log now says what it ran in: the browser (or the Obeya app), whether its window was in use, and when the owner last moved, typed or scrolled in it. In Safari, the Obeya app and Firefox, a window in the background logs no stall: those browsers say nothing about where the time went, so a held-back window can't be told apart from a busy one. Chrome's pages still log from the background, since Chrome names the page's own scripts.
Before:
2026-11-05T07:27:11.028Z Obeya: a page (this machine, canvas repo) stalled for 1.4 s; meanwhile it took no click or key
After:
(nothing logged)
Obeya: a page (this machine, canvas repo; Safari 26, its window active, last used 2 s before) stalled for 1.2 s; meanwhile it took no click or key before it: usage 1.9 s before; usage 1.2 s before
Obeya: a page (this machine, canvas repo; Chrome 149, its window active, not used since it loaded) stalled for 1.3 s; meanwhile it took no click or key 1.3 s TimerHandler:setTimeout: (anonymous)
On 3 November, from 10:48 on, the display was on and only the user was idle. If the Obeya window still had the focus then, stalls like those will still be logged, now with "its window active" and how long the page had gone unused. The next look at stalls.log will tell whether that happens.