Explore possibility of sampling more often in "max bytes used" perf tess
It is my understanding that sampling the memory more often leads to more reliable "max bytes used" perf results. Doing so makes the test run slower, however.
Now that we have a stable CI infrastructure with sufficient resources, is it worth increasing the sampling rate during validation? This would lead to fewer spurious perf failures, I think.
So, two questions:
-
Am I right here? That is, would increasing sampling actually help?
-
What is the cost of doing so always?