These five could not go in the previous commit because they hold PLN's own
uncommitted work from this afternoon's recording session, mixed in with the role
alignment. Committing them anyway, because durability beats a tidy diff: once they
are in git, a stale Pulsar buffer overwriting one of them shows up as a diff
instead of vanishing silently (#95, the buffer-clobber trap).

## What is HIS in this commit

- **bombe_dj**: he resolved the d1 button overflow BY HAND, exactly the way #54
  proposes. The three lines phase 2 had commented out with a FIXME are back, folded
  onto d1's own controls: the sample swap moved to a KNOB (`midiOn "^29" (# "kick:4")`)
  and the kick variants merged onto `^41`. That is a manual gSel, and it is better
  than what the FIXME asked for.
- **the_revolution_will_be_sampled**: he fixed the d8 inversion himself, restoring
  `midiOn "^92" (ply ...)` / `midiOff "^60" (mask ...)` before any tool got there.
- **vague_de_crime**: a `# cps (130/60/4)` gesture parked as a comment and another
  moved onto ^30; d1's commented overflow restored onto ^41.
- **do_it_right**: a `dr` typo removed, a d5 gain nudged 1.3 -> 1.4, a `-- TODO:
  Move to d6?` note, and an extra d5 ply gesture.
- **take_5_drops**: the `31` -> `21` typo in d12's `n` pattern, one `>|` -> `|>` on
  d5, gains adjusted.

## What is MINE

The button-role alignment only — `"^NN"` token moves, per the two commits above.

## One thing worth recording about the validation

bombe_dj's d7 exposed a limit of verdict-equality checking. The swap rewrote
`midiOn ("^59" - "^91")` to `midiOn ("^91" - "^59")`, and subtraction is not
commutative — that LOOKS like an inverted gesture. It is not: all three references
in the block flipped together (the two operands and the outer/inner `midiOn` pair),
so it is a consistent relabeling and the behaviour is identical with the two
physical buttons exchanged.

But silent-eval could not have told me either way. Both CCs seed to 0, so
`0 - 0 = 0` whichever order they are in, and the default branch is untouched by
construction. **Verdict equality at the seed proves nothing about behaviour under a
knob press.** It is the right gate for "did this rewrite silence anything at boot"
and the wrong gate for "does this control still do what it did". The second
question needs pv-at (#74), or an ear.