## Notes for the highlight/audit stage

- **No agenda was available** for this meeting (see intake note in `meeting.md`),
  so there is no roll-call document to close the candidate set against — unlike
  prior bodies in this archive where a published roster let role-only labels be
  upgraded to `probable`/`confirmed` by elimination. A vault search of
  `Ledger/Entities/People/` and prior BZA staging turned up no existing roster
  for this body, so none of the five names read into the record during roll call
  (Bill Treadway, Donnie Jeffers, Jim Wallace, Jason Saavedra, John Collette —
  all voiced by the `<other>` merged cluster or by `speaker_4` reading the list,
  never self-identified against a specific label) could be pinned to any of the
  nine substantial-speech labels below. **No name is invented to fill that gap.**
- A cross-meeting note: `Staging/2026-09-02-area-plan-commission/speakers.md`
  independently resolves a `speaker_46` to "Sydney, staff (planner)" at
  `probable`. That is corroborating, non-anchored context for this meeting's
  staff-report giver being a real county planning staffer named Sydney — but it
  is a different meeting, a different diarization run, and a different label,
  so it does not license upgrading `speaker_16` here. It is left `role_only`
  for the reason below, not for lack of a name to guess.
- **`speaker_9` and `speaker_45` are almost certainly the same person** (the
  chair) split across two diarization labels — both run agenda procedure in the
  same voice and cadence, with no other chair-like speech between their spans.
  Flagged via `merge_with` rather than silently merged.
- **`speaker_16` carries contradictory evidence** — it delivers what reads as
  the staff report itself (matching the chair's introduction "a staff report
  from Sydney" at [t=31]) but also thanks "Sidney" by name in the third person
  at [t=144]. A self-reporting Sydney would not thank themselves by name mid-own-turn;
  this is diarization contamination (the chair's brief "thank you" merged into
  the staff cluster). Rather than pick the more frequent reading and assign the
  name "Sydney," this is marked `role_only` (staff), per the diarization-
  contamination rule.
- **`speaker_42` shows the same kind of blending** — a member's motion language
  immediately followed by chair-style vote-tallying language in one label. Left
  `role_only` (board member) rather than guessed.
- **`speaker_27` (petitioner) is not named "Joey Floyd"** despite that name
  appearing in the staff report immediately prior ([t=52], "the owner is Joey
  Floyd"): the staff report names the *property owner*, and the chair's
  petitioner exchange that follows never confirms the person answering is that
  same owner rather than an agent, contractor, or family member. That is an
  inference about the case file, not a naming event about this speaker, so it
  does not clear even `probable`.
- 11 sparse labels of the original 20 are merged into `<other>` in the source
  transcript and are out of scope for this roster per the task instructions —
  no entries were fabricated for them.
- **Zero `confirmed` or `probable` assignments this meeting.** Every substantial
  label resolves to a function (chair, clerk, staff, petitioner, board member)
  but not to a name. If a BZA roster or agenda surfaces later, this note should
  be revisited before any `Ledger/Highlights/` note for this meeting attributes
  a vote or motion to a named individual — per the standard, that requires
  `confirmed`, which nothing here reaches.