XAMPPRocky

I can understand that, I'm always weary of adding something as option without a compelling reason to do so, as it adds another variation to test, and adds cognitive overhead...

@markmandel Sorry to clarify, I said health, but really I meant *readiness*. So the Quilkin would be marked as Unready and thus wouldn't be restarted but would be opted out...

> We should triple check that the proxy doesn't get it's connections disconnected if it moves from ready -> not ready. I'm actually not sure what happens. I have a...

I've also been seeing this, I'll investigate further on Monday, I think the retry logic in the proxy isn't retrying.

> This is the last log from the xds server before it restarted That makes sense why it would cause the management server to restart, the bigger issue IMO, is...

@luckyswede I can't replicate it by just deleting it in GCP. Can you set `RUST_LOG=info,quilkin=trace`, and provide the full log for both services?

@luckyswede Thank you so much, these are very helpful. The proxy not retrying seemed to be a red herring, it seems to be the xds server not providing an up...

Okay I think I've found what the issue is, but I haven't figured it out how to fix it and need to implement some other parts which are higher priority...

@luckyswede You should look for the same thing as last time, a proxy not having an update so it's missing a token and dropping packets, and you see that token...

Well this does eliminate one possibility (which is that the change requests notifications were being sliently dropped). It seems like `rx.changed()` is not enough, maybe we need to track individual...