Stu Hood

I've updated the title, because this also applies to stragglers in linting itself (and the PEX-repository case goes away in `2.11.x`). One other potential concrete improvement would be to speculate...

Another random thought on this: the current thinking that: https://github.com/pantsbuild/pants/blob/14e01228487f5a757941cf04756b09af9d6dd799/src/rust/engine/process_execution/src/bounded.rs#L265-L266 ...is only partially true: we could actually look at processes queued to acquire the semaphore. It's only processes which are...

After a failure in the foreground like that, is there anything interesting in `.pants.d/pants.log`? We attempt to set up a logger that will render more information for signals like that.

Hm... I wonder if the `poll_workunits` aspect is relevant. That's a pretty isolated codepath running in a background thread.

A user helpfully reported the following backtrace from one of the cores, which also points to `poll_workunits`: ``` [Current thread is 1 (Thread 0x7fe99a924640 (LWP 3724751))] (gdb) backtrace #0 raise...

Directly panicing in the relevant function is properly caught on macOS, but I have not tried on Linux. Have reported this as https://github.com/PyO3/pyo3/issues/2102, but there is more to do on...

Regarding the MDB abort: have opened #14214 to resolve #14115. As to the `poll_workunits` panic: I attempted to report it upstream: https://github.com/PyO3/pyo3/issues/2102, but my attempts to add debug output to...

This may have been addressed by #16093, which was fixed by #16120. Are you still able to reproduce?

> #13464 fixed this at the process level (doesn't cache the process -> result), but we still _memoize_ the `rule` run. Any `@rule` which depends (transitively) on the run of...

~Hm, but also: this is unlikely to have anything to do with process or `@rule` caching, because it is an `InteractiveProcess`. `InteractiveProcess` is never cached, and is invoked directly by...