This reverts part of commit 4a2fe0cf971766930019d6a52f31e32b818dae2f and fixes the test instability in a better way as tipped off by @AmeliasCode in this comment and specifically shared by @gwy15 in this comment.
It seems the problem was that after the first test finished, the runtime that 'owned' the Mutex or channel was shut down, breaking it for all subsequent users. By changing our tests to all use a shared runtime, the channel isn't killed and can therefore work in subsequent tests.
This reverts part of commit 4a2fe0cf971766930019d6a52f31e32b818dae2f and fixes the test instability in a better way as tipped off by @AmeliasCode in this comment and specifically shared by @gwy15 in this comment.
It seems the problem was that after the first test finished, the runtime that 'owned' the Mutex or channel was shut down, breaking it for all subsequent users. By changing our tests to all use a shared runtime, the channel isn't killed and can therefore work in subsequent tests.