feat(610): split the decision corpus into one YAML-frontmatter file per record

168 records -> docs/decisions/records/<area>/<topic>.md (163 active, 23 dirs) and
docs/decisions/archive/<area>/<topic>.md (5 archived). The filename IS the key,
so one-active-record-per-key becomes a filesystem property rather than a
validator check, and supersession becomes a `git mv`.

WHY: the monolith was a concurrency problem before an aesthetic one. A
3,900-line append target made parallel sessions collide -- PR #605 and PR #614
both hit append-vs-append conflicts during routine rebases, and hand-resolving
those inside the corpus is exactly the operation the rationale-rewrite guard
exists to police.

HOW IT IS VERIFIED: a ~170-file diff cannot be meaningfully read, so correctness
does not rest on reading it. The parser was taught BOTH formats first, so the
body-diff guard parses the old form at the merge-base and the new form at head --
the migration validates itself, no bypass. The proof is a field-level equivalence
harness: 168 records before and after, zero lost, zero gained, zero field
mismatches, zero rationale bodies differing. Reviewers should scrutinise the
harness; it is the actual evidence.

What measuring caught that reading would not have:

- ~500 lines sit OUTSIDE any record -- decisions.md's lifecycle schema and each
  topic file's preamble, mostly the only copy. Source files are kept and
  stripped, never deleted. They also cannot be filed per-area: topic files hold
  several areas and 4 of 23 areas span several files.
- Archive discovery was a non-recursive glob; after the split it found ZERO
  archived records, surfacing as four bogus "supersedes points to unknown key"
  errors rather than an obvious failure.
- ~32 live docs point into the corpus BY DATE, which the split dangles. Each
  stripped file now ends with a generated "Records formerly in this file" index,
  which also rescues the identical breadcrumbs in old issue comments.
- decisions.md's "In this file:" list was 97 same-file anchor bullets that the
  split makes WRONG, not merely stale. Dropped; the generated index replaces
  them with links that resolve.

The equivalence harness now runs against a checked-in FIXTURE, not the live
corpus. The earlier version migrated the real tree, which made it a one-shot:
the moment the migration landed there was nothing left to move and the tests
failed for reasons unrelated to the code. A fixture keeps them testing the
SCRIPT rather than the repo's current state.

Keys preserved verbatim, warts included: `sched` (12) and `scheduling` (1) remain
two directories for one concept. Renaming a key is not a move -- it changes
identity, breaks the equivalence proof, and invalidates MemPalace's per-key
drawers. Taxonomy normalisation is separate work.

refs #610

docs/decisions/archive/scan/musicvideo-reconciliation.md

@@ -0,0 +1,46 @@
+---
+key: scan.musicvideo-reconciliation
+title: 2026-07-20 — `JellyfinMusicVideoLibraryScanner` reconciles by library-scoped path diff + hard delete, not server itemId soft-trash (#494)
+status: superseded
+since: '2026-07-20'
+supersedes: none
+superseded-by: scan.musicvideo-server-identity@2026-07-25
+rule: (superseded) `JellyfinMusicVideoLibraryScanner` reconciles removed music videos by a library-scoped local-path diff plus hard delete (`TrashMissingMusicVideos`), not the server-itemId soft-trash pattern the other media-server scanners use, because music videos carry no server identity.
+signals: 'music-video trash sweep, path-based identity, cross-kind safety, path-keyed identity, empty-fetch guard reuse, remove-stale+add-new dedup · paths: `JellyfinMusicVideoLibraryScanner.TrashMissingMusicVideos`, `FindMusicVideoPaths`/`DeleteByPath`, `IMusicVideoRepository`, `MediaServerReconciliationGuard` · issues: #494, #477, #488, #496, #500'
+mechanics: '`ScanLibrary_Should_Not_CrossDelete_Movie_Or_Show_Sharing_The_LibraryPath`; `ScanLibrary_Should_Not_Sweep_When_Jellyfin_Returns_Zero_Items`; integration tests extending the #488 harness. #500 — when mirroring the remove-stale + add-new idiom, dedup the incoming set on **the same key its add filter compares** (the filter is materialized before the loop mutates `existing`, so duplicates both pass): `Name`, `Guid` for guids, and for Plex `Actors` an artwork-preferring dedup shared with the remove filter (whose key is `(Name, artwork-presence)`). Remaining un-deduped copies of the idiom: #600. Superseded by `scan.musicvideo-server-identity` (ersatztv#496): music videos gained a `JellyfinMusicVideo` ItemId/Etag identity, so the path diff + hard delete became an itemId diff + soft `FileNotFound` trash.'
+---
+
+The Jellyfin music-video scanner did add/update only — a music video removed on the Jellyfin side lingered in
+ErsatzTV forever and could still be scheduled. It now runs a trash sweep at the end of `ScanLibrary`
+(`TrashMissingMusicVideos`), mirroring the `MediaServer{Movie,Television,OtherVideo}LibraryScanner` "gone
+upstream ⇒ remove" pattern but with a deliberately different identity function, because music videos lack the
+media-server identity those base scanners rely on.
+
+- **Identity is (LibraryPathId, path), not server itemId.** The base scanners diff `GetExisting*` (keyed by
+  `MediaServerItemId`) against the incoming server item ids, then soft-trash via `FlagFileNotFound`. Music videos
+  have **no `JellyfinMusicVideo` entity and no `ItemId`/`Etag`** — the scanner is a standalone
+  `IJellyfinMusicVideoLibraryScanner` that injects the *local* `IMusicVideoRepository`, which offers no
+  itemId-keyed existing-set or flag seam. So the sweep diffs the **local path** set instead: existing =
+  `FindMusicVideoPaths(libraryPath)` `.Except` the incoming items' replaced local paths, then hard-deletes the
+  remainder with `DeleteByPath` + `IScannerProxy.RemoveMediaItems`, and cleans now-empty artists with
+  `IArtistRepository.DeleteEmptyArtists`. Hard delete (not soft `FileNotFound` trash) because there is no
+  per-item FileNotFound seam on this path and the issue's Done-when is "removed".
+- **Cross-kind safety is a property of the queries, not the media kind.** `MediaItem` is TPT with `LibraryPathId`
+  on the abstract base, so a Movie, Show and MusicVideo can share one `LibraryPath` (a mixed Jellyfin library).
+  Both `FindMusicVideoPaths` and `DeleteByPath` filter `LibraryPathId` **and** join the concrete `MusicVideo`
+  table, so the sweep can only ever see/delete music videos — a Movie/Show under the same `LibraryPath` is
+  invisible to it. Pinned by `ScanLibrary_Should_Not_CrossDelete_Movie_Or_Show_Sharing_The_LibraryPath`.
+- **Reuses the #477 empty-fetch guard.** The sweep is gated by `MediaServerReconciliationGuard.ShouldFlagMissing`
+  — a successful fetch that returns zero items (server mid-restore / transient) is indistinguishable from a real
+  emptying, so the whole-library wipe is refused and logged. Pinned as a negative control by
+  `ScanLibrary_Should_Not_Sweep_When_Jellyfin_Returns_Zero_Items` (removing the guard flips it red).
+- **Known limitation (deferred to per-library identity).** `MusicVideoRepository.GetOrAdd` dedups a path
+  **globally** (no `LibraryPathId` predicate), so a file served by two libraries with overlapping local paths is
+  a single row owned by whichever library scanned it first. If that owner later stops reporting the file while
+  another library still serves it, this sweep removes the shared row. A proper fix needs per-library music-video
+  identity (a `JellyfinMusicVideo` etag entity + migration) — the issue's "option 2 / fold into the base
+  scanner" refactor — tracked as #496.
+- **Tests.** Integration tests (real `ArtistRepository`/`MusicVideoRepository`/`LibraryRepository` over in-memory
+  SQLite, extending the #488 harness) pin removal, empty-artist cleanup, cross-kind safety, and the empty-fetch
+  guard. Proven non-vacuous: all four fail against the pre-fix scanner except the guard control, which only
+  earns its keep once the sweep exists.