Alexey Zapparov

Well I didn't. But after thinking sometime how problem could have disappeared, I realized where it was happening. And it turns out that it was you who fixed the issue......

In other words. This should be considered as fixed/closed, yes. :D

Integration is on webmock's side, thus having CI for this on our end seems a bit weird to me. Thus I believe such regression test should be added to webmock's...

Basically they just have a different shims and specs for those. :D In any case, I guess it would be awesome to start with test case that will reproduce this...

Without monkey-patches we will have to provide a global middlewares (same way as #499). And this will be even more brittle, as for WebMock shimming must be as close to...

Totally agree!

Yeah. Just can't find enough time to sit down and rewrite this thingy...

Yeah, that's unfortunate. Redirection and persistence atm are conflicting when used at the same time. That should be solved by introducing `HTTP::Session` class that will keep pool of persistent-clients per-URI...

https://github.com/httprb/http/pull/651 introduced the regression. Will work on this.

FWIW `HTTP::Client` itself isn't thread-safe: https://github.com/httprb/http/blob/6240672bc84b23339fc9a9878040fcb45db78fb5/lib/http/client.rb#L71-L76