Travis Downs

For `io_uring` the benefit is much less on that specific test, probably because the overall io_uring behavior is quite different as it does not do "poll-aio=1" behavior: it does to...

I wanted to add that for all backends we will in some cases (the case where we get all the way to the underlying backend "wait for events" function) make...

@avikivity or @xemul let me know if there's anything I can do to help move this along or also if it's DOA etc.

@xemul any hope that you can take a peek at this one? In our experience this is a big improvement to the accuracy and hence usefulness of the steal time...

https://github.com/scylladb/seastar/commit/b6fac76315e4ddf0dac9ba43823ee2e908bc1ea4 is purely a rebase to the tip of master.

@xemul all feedback addressed or at least responded to.

Thanks for the review @xemul !

@jcsp - right, I just noticed that and came back here to say that I'm not certain about this one anymore. The key is that, as you mention, `raise_on_crash` is...

Changed to sev/low since it only happens when debug logging is on.

Maybe the same as #9286 .