Document what Engine DJ does to a converted library

[jrgutier]

Docs and comments only — no behaviour change.

Two things measured on a real 3,673-track stick that were previously undocumented, and both of which look like defects if you don't know about them.

1. Opening Engine DJ can merge its desktop library onto the stick

Convert → clean verify → open Engine DJ → verify again:

track library: playlist_count: expected=45 actual=48

naming playlists that exist in no rekordbox playlist. Controlled repeat with Engine's desktop database cleared, identical conversion both times:

Engine desktop library	Playlists after opening Engine	Rows not written by us
Populated	48 (3 added, ids past our allocation)	6
Cleared	45 (none added, no row changed)	0

So the extras were Engine syncing its own desktop library — a legitimate feature, not corruption and not something rb2engine wrote. Their lastEditTime values predated the conversion, consistent with rows that already existed on the desktop side.

verify was correct to report it. Troubleshooting now explains the cause, gives the one-line SQL to tell Engine's rows from ours, and says plainly: verify right after convert, before launching Engine DJ.

The discriminator matters — we pin lastEditTime to the epoch on every playlist row we write, so anything with a real timestamp came from Engine. That survives Engine reassigning ids, which the id watermark does not.

2. Engine populates currentPlayedIndiciator itself

Diffing a pristine conversion against the same database after Engine opened and closed it, across schema, pragmas and every cell (blobs by hash):

Information.currentPlayedIndiciator:  0  →  1698144667125441751

That was the only change. No schema objects added, no pragma changes, and Track, PerformanceData, AlbumArt, Playlist, PlaylistEntity all byte-identical.

Recorded next to _DETERMINISTIC_PLAYED_INDICATOR so the question "should we write that field too?" has a documented answer: no. Engine sets it on first open, and writing an opaque constant would forfeit byte-identical rebuilds for nothing.

Verification

• 704 tests pass, ruff + mypy clean
• Stick re-verified end-to-end after a fresh rekordbox re-export: 0 discrepancies, 3,673 matched

🤖 Generated with Claude Code

https://claude.ai/code/session_01V3aF4y4GdsuJta2bfYUZ8w