This is because retry-on-failure is enabled upstream. This is used for most upstream test suites.

So the fact that it's green, isn't a bug. However, it would be far better to not rerun tests on failure. Unfortunately I don't think our tests/code/infrastructure is reliable enough to simply turn this on (see recent chromium-dev thread about flake and retries). It also happens on ios-simulator with unit tests:
https://build.chromium.org/p/chromium.mac/builders/ios-simulator/builds/25899/steps/ios_chrome_unittests%20%28iPhone%206s%20Plus%20iOS%2011.0%29

Here are a couple of options - I think we should go for #2, though it's the harder one.
1. Rerun failed tests downstream - this will make it consistent, but mask failures.
2. Turn off retry-on-failure upstream - this is a much harder thing to accomplish, but is the direction I think we should go in.

Here retry-on-failure really mess up because the test only fail if run as part of the full suite. If run individually (or skipping all tests coming before) they pass. I think this is because they depend on some global state that is messed up by another test run before :-/

huangml: Please reassign if this isn't yours.

I think CredentialManagerEGTest fails because "--enable-features=SlimNavigationManager" is not passed in to integration_egtest suite. Same test with the flag is working fine on chromium.fyi.  But not sure why they'll pass on retries.
https://logs.chromium.org/v/?s=chromium%2Fbb%2Fchromium.fyi%2Fios-simulator%2F13049%2F%2B%2Frecipes%2Fsteps%2Fios_chrome_integration_egtests__iPhone_6s_iOS_11.0_%2F0%2Fstdout

Can it be related to issue 788201's root cause (yet to be determined)?