[hey-research-lab]

What was wrong

harnesses/token-quote-coverage probes every token with "1 USDC -> token" (README, step 3). On BNB Chain the input is USDC at 0x8AC76a51cc950d9822D68b83fE1Ad97B32Cd580d, whose decimals() is 18 (read on chain). Four adapters send a flat 1000000 base units on every chain:

• cmd/monitor/kyberswap.go:53 amountIn=1000000
• cmd/monitor/paraswap.go:52 amount=1000000, and :54 srcDecimals=6
• cmd/monitor/odos.go:64 "amount": "1000000"
• cmd/monitor/relay.go:62 "amount": "1000000"

That is 1 USDC on Base and Robinhood Chain (6 decimals) but 0.000000000001 USDC on BNB Chain, so the BNB cells measure whether a provider can route dust, not 1 USDC.

ParaSwap shows it directly. Live /prices on network 56, USDC to CAKE (0x0E09...cE82) and to a thin four.meme token (0xC3A9...5c68):

request	CAKE	four.meme token
as sent now: amount=1000000&srcDecimals=6	400 ESTIMATED_LOSS_GREATER_THAN_MAX_IMPACT	404 No routes found with enough liquidity
1 USDC: amount=1000000000000000000&srcDecimals=18	200, route returned	200, route returned

The adapter counts any non-200 as a miss, so ParaSwap's BNB coverage is understated even for CAKE. KyberSwap and Relay still answered the dust amount in the same check, so their cells are not obviously wrong today, but they are not quoting what the bench says.

How it was verified

• Added cmd/monitor/quote_amount_test.go. It points each adapter at a recording http.RoundTripper (no network) and checks the amount sent for base, robinhood and bsc, plus ParaSwap's srcDecimals. Before the fix it fails on bsc for all four adapters (quoted "1000000" base units of the stable, want "1000000000000000000") and on srcDecimals="6", want "18". After the fix it passes.
• go vet ./... and go test ./... in harnesses/token-quote-coverage pass. The changed files are gofmt clean.
• No spec or site file changed, so pnpm validate/typecheck/build were not run.

What changed

• helpers.go: stableDecimals(chain) (18 on bsc, 6 elsewhere) and oneStableUnit(chain).
• KyberSwap, ParaSwap, Odos and Relay send oneStableUnit(token.Chain). Base, Robinhood Chain and Solana still send 1000000.
• ParaSwap sends srcDecimals from stableDecimals.
• Mobula (human amount 1) and Jupiter (Solana only) are unchanged.

Type

• Other (harness fix)

[Flotapponnier]

Thanks a lot for this one, and the timing is good. ParaSwap started validating srcDecimals against the token roughly 30 hours ago: from then on every BSC probe came back 400 "Source Decimals Mismatch" and ParaSwap's BSC coverage sat at 0 (it was around 60% the days before, while it was quoting 1e-12 USDC). With your change 1 USDC routes fine.

I pushed one small follow-up commit on top: ParaSwap now also rejects a wrong destDecimals ("Destination Decimals Mismatch"), and we were sending a flat 18 for every token, so I dropped the parameter (ParaSwap resolves it itself) and added a check to your test. Merged and deployed; the first cycle on the new build shows BSC ParaSwap quotes succeeding again.