Discussed Nov 14 with team and Justin Pierce. Could we collect full samples of prom CPU timeseries data and push to bigquery, to be used to correlate E2E tests that may be causing it (currently represented as intervals in bigquery).

Worth examining if the questions we need answered can already by via the CPUMonitor intervals that exist today. (any node over 95%) Justin made the case for how powerful it is to have the full data though.

Lot of prior art laying around, the CPUMonitor intervals code will have a useful query.