performance_browser_tests frequently breaks by unintended changes, and this breakage is not being caught until this runs in the performance waterfall. This causes a lot of churn because then the tests must be manually disabled, the breakage manually RCR'ed, and then all manually fixed. Furthermore, in the meantime, all performance regressions are not being caught.

Instead, the CQ and regular waterfall should run these tests to allow our automated infrastructure to keep them green before a breaking change can be committed.