caio

Hah I half expected this to come up :grimacing: The problem is that custom broadcasts use [the same mechanism](https://github.com/caio/foca/blob/main/src/broadcast.rs#L101) used for cluster broadcasts: it's a pretty dumb sorted array. SWIM...

Sweet! I think I over-promised a bit on the general purpose broadcasting feature :sweat_smile: I think when optimizing these sort of broadcasts you might end up wanting to do some...

That's right, foca doesn't try to do anything smart. It simply buffers a broadcast up and tries to send it `Confix::max_transmissions` times; it doesn't keep track of who it sent...

closing stale issues that seem resolved. feel free to reopen

Yeah, it's expected. I should definitely make this clear in the docs. Thanks for the nudge! I think re-announcing to the bootstrap list is a good approach (it's kind of...

Ooh interesting! My guess is that this node started declaring members as down by itself (machine is thrashing, network i/o grinds to a halt, probe cycle "fails") and then when...

> That could work. Is that the equivalent of a "non-fast" rejoin? Kinda. Every valid message is interpreted as a join [^1], the interesting bit about Announce is that it...

Ooof that's pretty weird! But I think I know what's up with the low cluster size. If you revert the config change and restart that just one node I suspect...

Hah I suspect you were looking at the side effects of an unclean restart: When an instance leaves the cluster without notifying it (i.e.: without `Foca::leave_cluster`, for example), it takes...

Ooh that's interesting! Are you able to dump `foca.iter_active()` or perhaps the per-address counts? Would be helpful to know if this is a cluster-wide problem and many nodes have multiple...