Wanted to discuss this before opening a PR: is everyone okay with a bulk conversion from the old-style ASSERT macro to the new-style ASSERT_EQ, ASSERT_NE, etc. macros?

It'll cause a bit of churn in test/ but it should make debugging tests a lot easier.

Currently we use a pretty naive removal of testing dependencies for SwiftPM by setting an environment variable. We should be able to use Rocket and update our Rakefile script to just remove testing dependencies on each release, and keep it in the development process thanks to that.

The idea came from this PR that I saw RxSwiftCommunity/RxOptional#83, which sounds reall

If you want to work with a complete socket address type including port, we have a helper in SocketAddress. However, the moment you want to talk only about an IP address we force you to hold an in_addr or in_addr6 structure: hardly the friendliest versions of these data types.

We should provide a helpful wrapper IPAddress type that can be converted into those types as needed. This would

It would be useful to be able to toggle/flag into a process-level mode to group by pid and display more process-level information like the full command line.

• For the "Utilization by process name" window, we could group by pid rather than process name so that it's easier to find specific heavy network resource consumers.
• For the "Utilization by connection" window, including the pid + cmdlin

What should happen if deliver_cancel raises an exception? In the current implementation, the shielded cancel scope will prevent it from propagating until the process exits, which might take a while if the crash occurred before signaling the process in any way. Maybe on exception from a user-specified deliver_cancel we should call the default deliver_cancel to kill the process? Or just kill() since

We currently only test our autoscaling behavior for concurrency and rps. We should add an e2e test that also asserts we properly scale on `