The wall
The rule
The in-office voting record feeds the Monkey and the Llama. Nothing else does. Ever.
VOTES, AUTHORSHIP, BLOCKING -> Monkey Award, Llama Award
everything else -> Flamingo Award
"Everything else" means transcriptions of interviews, background research, resumes, education, public statements, and qualification assessments. None of it touches the two voting-record awards, and the voting record does not touch the Flamingo.
Why the wall exists
Because the two kinds of evidence have completely different reliability, and mixing them would let the weaker kind contaminate the stronger.
A vote is a fact. It happened, it is in the public record with a date and a citation, and the subject cannot reasonably dispute that it occurred. That is the firmest ground the movement stands on, and it is worth protecting absolutely.
An assessment of somebody's interviews is a reading. A careful one, against a rubric published in advance, applied consistently — but a reading. It is legitimately arguable in a way a roll call is not.
Let the second into the first, and the Monkey Award stops being "here is what this person did, here is the citation" and becomes "here is what they did, plus our impression of them". The moment that happens, the strongest claim the movement makes is only as defensible as its weakest input.
Both can render at once
Somebody who held office and is now a new candidate carries both: an award percentage from the record on one side, a flamingo state on the other, side by side, neither leaking into the other.
How it is enforced
Not by remembering. The separation is a structural property of how the awards are computed — they read from different sources — and it is checked in the build, in the same spirit as the movement's other hard rules.
A rule that lives only in a specification is a rule somebody optimises away under deadline.
The practical consequence for readers
When you see a Monkey or Llama percentage, you are looking at arithmetic over cited acts, and you can reproduce it with a calculator.
When you see a Flamingo, you are looking at an assessment against a published rubric, with a percent-known figure telling you how much of the material we actually have.
Those are different kinds of claim, and the product refuses to present them as the same one.