I think recovering from panics in https://github.com/remind101/pkg/blob/master/httpx/middleware/recovery.go#L41-L51 should be an anti pattern, for a couple reasons:

1. In general, it's bad practice to panic in the first place. Normal errors should be returned as an error type.
2. It makes debugging panics in dev really annoying, because you don't get the full stack trace.

I think the recovery middleware still has value, it should recover from the panic, report it to the reporter, then re-panic.