Re mos-0920. I have to correct a line from my 02:14Z message before I record deviation 3. It said the token limits had already been read through the human Console path at W9b console-check. That's wrong. console-check.sh has no limits check (grep finds no "limit" in it). Its W9b PASS lines were markdown-preview, /api/config, project.files, project.file north-star and loopback save 403. The last sentence of your deviation 3 text came from my claim, so I won't record it as you wrote it.

What did read the limits in this window:
- W9i, mosaic limits as the tester at 01:44:25Z: launch concurrency limits (anthropic opus 4, sonnet 4, haiku 20) and "tokens": [], meaning no token limit rows are set.
- W10b: the tester Settings page sent GET /api/bus/limits twice through 7333 with no refusal (bus.js:816). The isTester branch at :817 then drops the rows without rendering them.
- No step rendered a Token limits panel. The human Settings path (bus.js:821, tokenLimits) didn't run in this window, and I'm not adding a human Console read.

Proposed deviation 3 wording: "Board and Settings read. The tester Settings has no Token limits panel, by design at bus.js settingsView isTester branch. No daily quota entry shows. Token limits read as the tester: W9i mosaic limits shows tokens [] (none set), and W10b's tester page GETs /api/bus/limits without refusal. No step rendered the human Token limits panel." I'll put that in W10.txt along with this correction and the line you dictated. If you want different words, I'll append them as a new entry.

Going ahead now with the 7332 403 check and P4's tunnel read, since neither depends on this wording.
