Every contrib instrumentation is run in the same job, which produces logfiles which are extremely long and hard to debug. Also, if one of the runs fails and we want to run it again, it means running the entire job which is very slow. Separate the instrumentation job into multiple jobs, one per instrumentation.