phiresky

One of the federation tests is failing again (on the second pass) :/

> cargo test duration seems to have increased from 5 minutes to 15 minutes which is a huge increase. Any idea why? There's a fair amount of forced-sequential tests in...

@Nutomic yeah that inconsistency is due to [this one of your patches](https://github.com/LemmyNet/lemmy/commit/087684658a5d2008dff3df54d9b6e0ebde1ce1bc). You tried to simplify the code and changed the numbers (not sure if intentional).  30s for the...

I'll revert the 60s to 1s for the latest_id cache, but we should keep the 30s recheck. > This way manual testing on prod servers will be much quicker (eg...

CI passes now. I've re-removed the double run of federation tests because it's too much of a chore to make them reliably run twice without resetting CI. @Nutomic still needs...

It seems I don't have merge permission anymore, so could one of you merge it please?

This _might_ be feasible, but I'd definitely be a lot more cautious about this than you or the OP in the socialhub, who seems to be writing from the perspective...

> Apparently our HTTP client doesnt keep connections alive Do you have a source on that? As far as I'm aware it does keep them alive and thus for replication...

Why is it non-trivial to solve? if the library provides an option to do it it's trivial to solve. the only problem is that by default it apparently uses HTTP2-pipelining...

> Is this [HTTP Response Streaming](https://gist.github.com/CMCDragonkai/6bfade6431e9ffb7fe88) It's [HTTP pipelining](https://en.wikipedia.org/wiki/HTTP_pipelining)  The docs you sent seem to be about responses, but here we really only care about request and request body