catching and tracing coredumps
I am never quite sure if supplemental services are crashing or not. We should probably have infrastructure backing for this.
What I would imagine is setting a core pattern on the host (/proc/sys/kernel/core_pattern)[1] that dumps into /tmp and encodes the binary path into the coredump file name. And then have the tooling right at the end of a build/test run look in that well known location for cores and if there are any run a quick lldb --batch --one-line bt --core foo binary
on it to do trivial attempt of tracing the crashes.