Twinned from https://github.com/cfsh/ninefold/issues/487 by tools/beads/import_github.py's reconcile pass.
From the playtest sitting 2026-08-04, F5 build ca7a756. Aedan's words:
> "I managed to create a weird bug. I'm looking at the keys pattern, but actually the roll I'm showing on the right is for drums. So I put in some notes, and I'm hearing note sounds, but I'm looking at a drums editor. There's a weird bug going on here."
Evidence: [capture-1](https://github.com/cfsh/ninefold/blob/6334c0a49699b4e877cc48eccc3e0913c06ae343/playtest-logs/2026-08-04-ca7a756/capture-1.png) and [capture-2](https://github.com/cfsh/ninefold/blob/6334c0a49699b4e877cc48eccc3e0913c06ae343/playtest-logs/2026-08-04-ca7a756/capture-2.png) — he shot it twice, 13 s apart, from the same state.
The journal corroborates it, and that is the useful part. Both captures record the editor's *model* state as:
``` zone=Notes pattern=hum sound=keys ```
while the *rendered* roll was, in his eyes, the drum grid. So this is not a mislabel — the view and the state genuinely disagree. Whichever is authoritative, one of them is wrong, and note entry was going somewhere he could hear but not see.
Corroborating detail: drums render as a fixed-lane grid rather than a pitched roll (PianoRoll._isDrum short-circuits the L2/R2 octave logic), so "looking at a drums editor" is a structurally different widget, not just different labels — consistent with the roll having been built for one sound and the pattern loaded from another.
Not deterministic in his hands — he hit it while moving between channels and patterns, in the sitting that also produced #453's screen-flip behaviour. The 8 focus transitions before capture-1 are all within zone=Notes on pattern=hum / sound=keys, including two visits to NotesChangeInstrument, so the change-instrument button is in the immediate history and is the first thing worth suspecting: changing the instrument while a pattern is open may rebuild the roll for the new sound without reloading (or without re-binding) the pattern.
Three anomaly: null focus-lost events also occurred in this sitting (01:53:16, 02:03:32, 02:03:33, all zone=SongGrid) — probably unrelated, noted so they aren't rediscovered.
Every other item from this sitting is polish. This one silently edits the wrong data: he added notes believing he was editing keys. Worth pinning with a probe assertion once the cause is known — the roll's bound sound and the loaded pattern's sound should be assertably equal.
Possibly related to the one-screen flip path (#453) — see #464's "uncovered flip-with-fill-panel path". Filed standalone rather than on #453 because it may well pre-date that work; whoever picks it up should check both.
None.
No comments.
