One of the system constraints unit tests is flaky. The resolver fails correctly, however, the reason given seems to change randomly. For more information:

https://github.com/operator-framework/operator-lifecycle-manager/issues/2564

This shouldn't be a blocker for shipping. I think this is the "normal" behavior of the resolver right now. Though we should understand what the source of stochasticity  and see if we can mitigate it. The resolver should be as deterministic as possible.