"Squash failure metrics to unique categories for each build" option can be useful when a single failed build should count towards each cause category exactly once. This is particularly useful when combined with the "Treat failed test cases as failure causes." If numerous tests fail, "squash failure metrics" will prevent those categories from being incremented for every single failed test, which can skew metrics significantly.