With the changes added in #198, I've added some sanity checks to ensure that the benchmark finishes successfully.
We can extend this idea and add launching the benchmarks as part of CI (probably just with n=1) to ensure nothing is being broken. This would help with running benchmarks (as it may happen that a single benchmark failure will require the whole benchmark re-run) and increase the test coverage (ideally, all of the things done in benchmarks should be tested in integration tests, but there may be some gaps in coverage that I'm missing)
With the changes added in #198, I've added some sanity checks to ensure that the benchmark finishes successfully.
We can extend this idea and add launching the benchmarks as part of CI (probably just with n=1) to ensure nothing is being broken. This would help with running benchmarks (as it may happen that a single benchmark failure will require the whole benchmark re-run) and increase the test coverage (ideally, all of the things done in benchmarks should be tested in integration tests, but there may be some gaps in coverage that I'm missing)