docs(macos): record why the window suite loads its page once, at top level Measured on release run #128: the navigation case's in-suite `loadHTMLString(_:baseURL:)` sat at `estimatedProgress = 0.1` with `isLoading = true` for its whole 30 s budget on the x64 runner, while the identical load at top level commits in 0.3 s — five checks failed on a working web view and the arm64 leg passed with the same code. The page now carries the navigation case's elements and is loaded once with `loadSimulatedRequest` at an http origin, because the rule under test compares origins. `proveClickCloses()` runs last: it closes the window, and WebKit may suspend a web view whose window is closed.
docs: correct the release re-run claim and record the tar host-spec trap The stranded-release bullet said a re-run of `Release the app` "cannot fix it" because semantic-release sees the tag and exits. That is only true when `main` has not moved: the job fast-forwards to `main`'s tip first, so a re-run with new commits cuts a release from a commit the run never gated — measured on 2026-11-12, when re-running 43120058573 cut v0.29.0 from a red-gated commit. The Files panel's archive verbs get their own bullet: GNU tar reads a `C:\…` argument as a remote host, so the archive now arrives on stdin, and the tool lookup moved after the request's validation.
docs(releases): record the stranded-release recovery after a rejected notes push Releasing v0.28.0 stopped after semantic-release had already pushed `main` and the tag: the notes ref was rejected with `[remote rejected] (failed)`, which no repository rule explains (the only ruleset targets branches, and the same push succeeded by hand). The release was left with a tag, no npm publish and no GitHub release, and a re-run of the job repairs none of it because semantic-release sees the tag and exits without publishing. The bullet records the measured failure and the three-step repair that worked, including why an attach-only dispatch has nothing to download and what `version-code` the Android dispatch needs.
Revert "docs(releases): record the per-shell build gate" This reverts commit a2eb7d1. The gate it documented was withdrawn; the page goes back to describing a release that builds and carries every shell.
docs(releases): record the per-shell build gate The release pipeline now builds a shell only when its own version moved, and the two job-level guards that makes necessary are the part a reader would otherwise re-break: `shell-commit` and `app` must survive a SKIPPED build, and `app` must still refuse to tag when `shells` failed — a failed bump leaves every build skipped, and skipped is not a failure. Also records the consequence in the update probe: the resolver's scan window is load-bearing now that a release no longer carries every shell, so it scans 100 releases in one API request and the pages fallback walks two Atom pages.
docs(releases): record why a phone switch crashed and signed the user out The first version of the console's Server chip on Android and iOS shipped two faults, both now measured rather than inferred. `ServerHost` answers every bridge command on its own worker, so `onActivate` arrived off the main thread and touched the loading overlay and the WebView inline — `CalledFromWrongThreadException`, process dead, and a latched address that made `onResume` skip the retry. And both phones wiped the cookie jar on a switch, which is what ended the session: cookies are keyed by host and the session cookie is named per port, so the wipe protected nothing and signed the user out of every server in the list. The wiki entries for both shells now name the fault, the fix, the desktop behaviour that was right all along, and the emulator round trip that proves it.
docs: record the one-workflow-per-push and staged shell bump rules Two pipeline facts changed and each came with a measurement, so both belong here rather than only in the workflow comments a future reader may not open. A push to `main` now raises exactly one workflow. `ci.yml` and `e2e.yml` carried `push: branches: [main]` beside `release.yml` and re-ran the same reusable workflows on the same commit, and the duplicate pair had already disagreed: on 26333b1 the CI leg's `Gates / Checks (Windows)` failed on an `EBUSY` temp rmdir while the Release leg of the identical job passed. A shell version bump reaches `main` only after every shell build passed, on the staging branch `release/shell-bump`; `shell-commit` fast-forwards it once the APK, IPA, Windows, macOS and Linux builds are green. Committing first was the bug: a failed build left a bumped version on `main` for an artifact that never existed. The e2e page's claim that the browser layer runs on every push is corrected to what the pipeline now does: label-gated pull requests, nightly, and as a gate of the release run.
docs: record the phone server chip and the transport split The Android and iOS shells used to have exactly one way to change servers — the launcher shortcut's native settings screen — and the wiki said so. The console now draws the same server control the desktop shells have, from the drawer's title bar, so both shell pages say what the chip is, where it lives and why: measured at 390px with a password set, the chat header leaves the session title 89px and a 32px chip costs 40 of them, while the drawer row has ~158px spare. Also records the two facts a future reader would otherwise re-break: the launcher shortcut is still the only entry point when the saved address does not answer, and `hasShellBridge()` must stay distinct from `isDesktopShell()` or the titlebar reserves 5rem for traffic lights a phone does not have.
docs: split the manual into a wiki