Ayoub Mrini

> If I read this correctly you didn't implement early exit when the limit was hit; just added the interface so that someone else could do that. This seems wrong...

> I took a stab at implementing the limits inside merge.go I included those changes in the same PR as changes seemed minimal. Could you take another look @machine424 and...

> Even with no difference, I don't think this is working properly since the /os-floating-ips (compute) or /floatingips (network) API only returns data owned by the user's project. I also...

Thanks @alanprot, this lgtm ;)

Thanks for this! (I always appreciate creative tests) We do trim the queue before each `sendAll` iteration in `nextBatch`, if we couldn't send the alerts to any AM, we increase...

I had started reviewing this but never published my comments, I can take over if you want @bboreham

We may need to adjust the new https://github.com/prometheus/prometheus/blob/3b8b57700c469c7cde84e1d8f9d383cb8fe11ab0/scrape/scrape.go#L1625 as well. Do you think that adding the metric name (if available) makes sense here? Or is it better to have extra...

> In the context where I started working on this, called from Mimir, it already prints the full set of labels. Ok, we can delegate that to the error catcher/wrapper.

> @machine424 are you planning to review this? I wasn't aware of this, but I'm happy to review when it's ready ;)

Given all the possible exit points in `main`, it may not be easy to maintain/keep synced such code, but let's give it a try. For tests, if we could initialize...