test: retain verifier regression guards

[chhoumann]

Summary

Several plausible regressions passed the existing suites. Add twelve permanent cases from the independent verifier probes. They guard player fullscreen and CSS ordering, media resets and native seeks, settings sync, table-cell capture, transcript preservation and Find reset, and redundant decoding during own saves. Only test files change.

Verification

• npm run lint, npm run format:check, npm run typecheck, and npm run build pass.
• npm run test passes 2,613 tests in 124 files. npm run test:tooling passes 13 tests. npm run docs:build passes all internal links and 112 legacy fragments.
• npm run test:e2e from a fresh isolated vault passes 207 tests in 28 files on both Obsidian 1.14.4 and 1.13.7, each from a fresh vault.
• The unchanged Media Session journey passed 20/20 runs with 0 failures under two continuous SHA-256 load workers. The workers and runner shared CPU affinity 0 and 1. No wait or assertion changed for that journey.

Node 22.23.2. Obsidian 1.14.4 and 1.13.7, each with its matching installer, on headless Linux. The runner used only this worktree's isolated vault, podnotes-test-verifier-guards. Fixtures use http://podnotes.test:<ephemeral-port> with the player, media-reset, table-capture, and transcript feeds, 60-second silent WAV media, and the committed WebVTT cues. The Media Session journey uses podnotes-e2e-silence.wav, a 600-second local WAV in its test sandbox.

Each mutant ran alone and failed the named assertion. Every production file was restored byte for byte and checked against origin/master. The green results use the restored production code.

Item	Test name	Mutant	Red output	Green
V-472	the fullscreen button belongs to the video stage and hides while that stage is fullscreen	Hide on .pn-player:fullscreen instead of .pn-video:fullscreen	Expected none, received flex	Pass
V-472	mini-player component styles survive the shell stylesheet loading last in theme-light / theme-dark	Remove button from MiniPlayer's .pn-iconbtn selector	Opacity changes from 1 to 0.85; light text changes from rgb(34, 34, 34) to rgb(92, 92, 92)	Both pass
V-475	resume / play preserves 12 seconds before a reset element reports emptied	Replace holdsMedia() with the existence of session	Both receive 0, expected at least 12	Both pass
V-475	switching after a pending native seek saves and queues the ended outgoing episode	Reconcile only when playing the same episode	Expected queue to include Outgoing, received []	Pass
V-468b	blur keeps an external setting received after the focused edit was saved	Apply the stale input value again on completion	Expected Remote/{{title}}.md, received Local/{{title}}.md	Pass
V-471	falls back to the note editor when the cell's replaceSelection is not a function	Accept any cell object without checking its writer	TypeError: cell.replaceSelection is not a function	Pass
V-471	captures in the cell when the PodNotes leaf takes focus in the same evaluation	Ignore the pending active leaf	Expected destination editor, received note with a newly created episode note	Pass
V-478b	preserves the quoted feed and transcript URLs of a public source	Redact every transcript source URL, including public URLs	Expected https://show.test/transcripts/1.vtt, received podnotes-private-feed:Show	Pass
V-478b	switching episodes clears Find and shows every line of the incoming transcript	Remove the finding and query reset	Received three filtered lines and three marks instead of ten lines and no marks	Pass
V-480	does not decode its own saved document again during a settings session	Remove the equality shortcut in the data-file reader	Expected zero decodes after startup, received six	Pass

The coordinator can replay the exact mutation commands and inspect the raw failures in .obsidian-e2e-artifacts/test1/mutate.py, mutations.json, and *-red.log in this worktree. The loaded-run counts are in flake-watch.json. The full acceptance logs and decision trail are in the same directory. Both instances were stopped; the final dry runs report no PIDs.

Release impact

None. No production code, schema, API, URI, or migration changes.

Note

Add regression guard tests for capture, playback, transcript, and settings

• Adds unit and e2e tests for timestamp capture fallbacks: when the Live Preview table cell cannot replace its selection, and when PodNotes focus precedes the active-editor state update in commands.test.ts and table-cell-capture.test.ts
• Adds e2e playback coverage in playback-media-reset.test.ts: resume/play after a native media load reset preserves the paused position (12-15s window), and switching episodes after a pending native seek persists the outgoing episode at 12s as unfinished
• Adds e2e journeys: video fullscreen enter/exit via the fullscreen control, mini-player style stability under stylesheet reordering in both themes, transcript episode-switch clearing the prior Find query, and settings external-update retention when a focused field is blurred
• Adds unit tests for self-originated settings persistence without repeated decoding in main.persistence.test.ts and verbatim URL retention in transcript cue text in index.test.ts

^{Macroscope summarized 1a1dd08.}

[cloudflare-workers-and-pages]

Deploying podnotes with    Cloudflare Pages

Latest commit:	1a1dd08
Status:	✅  Deploy successful!
Preview URL:	https://0e8a1faf.podnotes.pages.dev
Branch Preview URL:	https://test-verifier-guards.podnotes.pages.dev

View logs