Michael Martin

Interesting! I am suspicious that there's something that is silently munging the balancer data. Purely speculating here, but that sounds slightly more likely than the bug being in the balancer...

Hey @marc-charpentier, sorry for the late check-in on this (we're all a little bit swamped at the moment prepping the next release). That test failure looks very similar to one...

This will probably require some updates to EE for WebSocket services since we deviate from the regular runloop handler code path in several places (not a problem--just making note it...

This has already been merged in EE master, due to the breakage it was causing, so it would be good to merge it in for 3.0 for consistency's sake (and...

Using sha1 seems like a pretty lateral change to me given that the field value was previously just part of the key, in plain text. Does LMDB encrypt key data...

@dndx ah that makes sense. To me the collision probability with CRC32 seems to put it out of the running in this use case. I assume md5 would raise the...

@fffonion I've switched to sha256. Can you take a quick look and merge if this is workable for you?

Hi @marc-charpentier :wave: This is a good find! Appreciate the debugging efforts and detailed report. Your reasoning on this seems sound to me--hopefully another dev will chime in to confirm...

Added the `legacy_hybrid_protocol` check to the test template. It didn't seem to break any tests, so :crossed_fingers: that this is good to go.

@chronolaw @Suika-Kong @hbagdi, any :+1:/:-1: on whether this should get merged for 3.0 or not? It's not a hill I would die on, but I think it's a worthwhile improvement,...