This is a follow-up from  issue 710710 : "If the first collected swarming task is expired, no remaining swarming tasks are collected"

The cascading failure in the bug was caused by the loop in https://cs.chromium.org/chromium/build/scripts/slave/recipe_modules/chromium_tests/api.py?rcl=424fd63dc75bf3fec55f76b98c932621c52577cf&l=286:

      for t in tests:
        try:
          t.run(self._api_for_tests, suffix)
        except self.m.step.InfraFailure:  # pragma: no cover
          raise
        except self.m.step.StepFailure:  # pragma: no cover
          failed_tests.append(t)
          if t.abort_on_failure:
            raise

Based on this code, any InfraFailure of any step will block the whole build, but I don't think we can guarantee a zero percent failure rate for  InfraFailure, especially given any SwarmingFailure will be an InfraFailure (see https://cs.chromium.org/chromium/build/scripts/slave/recipe_modules/swarming/api.py?rcl=424fd63dc75bf3fec55f76b98c932621c52577cf&l=668)

I think the loop here should be adjusted so that the build keeps going upon InfraFailure. 

What do other folks think?