Currently, the legacy spec runner loops over the runOn requirements creating test per requirement. It then depends on the filters to delete the tests that don't apply. With the semi-recent addition of a third topology "load-balanced" that doesn't appear in the legacy spec tests it's difficult to work with this style of test writing to force tests to run on serverless.

We should instead add the runOn requirement check to a beforeEach hook applied to all the tests from a given file. This allows us to control the runOn requirement much "closer" to the test and unblock with more granularity tests we need to run against serverless.