[Flotapponnier]

Both token-scoped providers shipped on Base and Solana, excluding BNB because
that column published the provider's own whole-second timestamp. #2794 moved
BNB onto the reference clock an hour later, which retired the reason, and the
stale reason stayed in two file headers, two published formulas and a
methodology bullet.

Measured before changing either table

45 s per chain, on the non-stable leg of the bench's USDT/WBNB PancakeSwap V3
pool:

provider	endpoint	result
Birdeye	/bsc/socket-optimize	825 trades, 825 with a hash
OKX	chainId 56	938 trades, 938 with a hash

BNB's reference is eth_subscribe logs on the bench pool address, so it is
pool-scoped like Solana: a token-scoped provider's off-pool emissions find no
match and are dropped. This adds a chain where the comparison is sound, and
deliberately does not touch the open question about Base, whose flashblock
reference is chain-wide.

Robinhood: ruled out by measurement, not by a chain list

provider	endpoint	result
Birdeye	/robinhood/socket-optimize	WELLCOME frame, 0 trades in 45 s
OKX	chainId 4663	acked, 0 trades in 40 s

Both on the pool address and on its non-stable leg, while Base and BNB on the
same code delivered hundreds. An invented chain name 404s where /robinhood/
does not, so the route exists and the chain is simply not indexed. Robinhood
keeps four of the six feeds.

Same ack-proves-nothing shape as the six pool-scoping attempts already
documented in the OKX header, which is why an accepted subscription is never
counted as coverage in this bench.

Three comments on main that contradict the code

• The Birdeye header told the next reader that "no TLS fingerprinting is
  involved and the uTLS machinery is not needed here", forty lines above the
  uTLS dialer that exists because it is and it was.
• errBirdeyeForbidden still asserted the egress-IP diagnosis that the Go toolchain was the fingerprint Cloudflare refused #2802
  retired.
• The OKX header gave Robinhood's absence as a fact read off a chain list.

The two guard tests

They fired on this change, which is what they were for. But they asserted the
conclusion (OKX must not cover BNB) rather than its premise (OKX may cover a
chain only while that chain reads a clock we hold). Re-derived as the coupling:
take bnb back out of referenceChains and the test now fails naming OKX as
the reason, which the unconditional pair could not do.

Published copy

• Cardinality bullet still said 4 aggregators x 4 chains = 48 series. It is 22
  aggregator-chain pairs x 3 regions = 66.
• Birdeye's coverage is now disclosed rather than implied by the presence
  ratio: 106 of 442 and 13 of 44 bench-pool swaps over two Base windows, about
  a quarter, so its percentiles describe the trades it reports and not the pool.

Gates

go build ./... and go test ./cmd/script/ -run 'Birdeye|OKX|Reference|Chain'
pass; pnpm validate reports 235 specs valid. Both formulas were over the 240
character cap on the first attempt and were shortened, with the detail moved
into methodology bullets.

Not in this PR

• The Base reference is chain-wide while the pool-scoped providers are scored
  on the bench pool alone. That affects OKX's already-published Base column and
  is a separate decision.
• The materialize worker clones dev at build time, so it needs a rebuild
  before either provider's new chain appears on the page.
• dev is 11 commits behind main on this bench: OKX on bench 001, and BNB onto the reference clock #2794 through the Go toolchain was the fingerprint Cloudflare refused #2802 all went
  to main directly and the fast-forward back to dev was never done.

🤖 Generated with Claude Code