Athos

#179 was closed in favor of #264, but the later was also closed :( Are there any other approaches we can try to prevent the reset overhead when retrieving connections...

> I assume it's important to you that the connection reset not happen at all on a "user" thread? I don't think this would be a problem for me. I...

An approach that I think would be equivalent to mine: - Starting the clean up when disposing the connection. - Adding it to the pool without `await`ing for the session...

Your considerations are spot on. What do you think about letting the user configure how they want to handle it? Would it introduce too much complexity? In my use case...

This would be perfect for me.

@glommer can you confirm that this is indeed a bug? It seems that [there is a test that expects this behavior](https://github.com/chiselstrike/chiselstrike/pull/1367#discussion_r888739210).

Is there a way to define a root/catch all endpoint? If there is such a way, then users can configure their default response.

I'm seeing a similar problem when trying to build [sqld](https://github.com/libsql/sqld/blob/main/Dockerfile). `cargo build` step runs from scratch, not using the builds generated by `cargo chef cook`. Dockerfile for reference: ```dockerfile #...

~@LucioFranco we should probably use 408 instead~ Probably not a good idea, I think 408 is reserved for when the client takes a while to send the body.

Let's not show broken databases to users instead. When they list databases, let's hide the ones without instances. When they create one that exists but has no instances, we just...