Concerns about reiability of GA

[CreepySkeleton]

We've been observing large fluctuation in benchmark results on essentially empty commits. It's apparently because GitHub Actions runs benches on different machines (come to think of it, it's sensible) with different load score.

It doesn't look like GA should be used for benchmarking to me.

[miparnisari]

GitHub Actions runs benches on different machines

Yep. In my project I solved this by keying the benchmark with the OS and the CPU:

https://github.com/openfga/openfga/blob/21e84f2dd763eadd0ce595ff470cfc5240f7d2d7/.github/workflows/main.yaml#L47-L53