In 210720 we skipped test stress/check-stack-overflow-before-value-profiling-arguments.js on ARMv7 which was failing on EWS. This failure is due to a specific condition that arose recently where the test is scheduled in the same rpi3 together with other memory-hungry tests that make this test crash. The same doesn't happen in the post-commit bot which happily passes the test. It's a shame to skip a test that passes simply because it happens to be scheduled with other memory hungry tests. We should possibly have a way to say that the test should not be scheduled on the remote with other tests so it runs standalone.