Intermittent fatal error from LeakSanitizer in CI (virtualized?) environments

[vadz]

We're running our unit tests in a build using ASAN under Travis CI and sometimes (not always, but often enough to be a problem in practice) they fail with

Tracer caught signal 11: addr=0x0 pc=0x7fdc954758b0 sp=0x7fdc8dfbfce0
==37877==LeakSanitizer has encountered a fatal error.
==37877==HINT: For debugging, try setting environment variable LSAN_OPTIONS=verbosity=1:log_threads=1
==37877==HINT: LeakSanitizer does not work under ptrace (strace, gdb, etc)

see e.g. this job.

Rerunning the job fixes the problem, but this is still annoying so any suggestions about fixing, or working around (e.g. detecting this and just trying to rerun the test once again) the problem would be very welcome.

Please note that we can't enable the recommended LSAN options because this results in too much output for Travis to handle and the job is killed because of it. Also, there are many similar reports, but all of them involve running in a docker container or similar, while we don't do anything like this and are running the test directly in the Travis VM.

[vadz]

This is unfortunately not specific to Travis CI, a similar issue happens with GitHub Actions too.