Android PFQ skips HWTests when Android container of new version is not available. You can see following logs in UprevAndroid stage when remaining stages are skipped:

18:19:18: INFO: Stable candidate found 4048825-r2
18:19:18: INFO: Previous ebuild with same version found and ebuild is redundant.
18:19:19: INFO: Found nothing to rev.
18:19:19: INFO: android_atom_to_build = None
18:19:19: INFO: Android already uprevved. Nothing else to do.

https://uberchromegw.corp.google.com/i/chromeos/builders/caroline-nyc-android-pfq/builds/126/steps/UprevAndroid/logs/stdio

But I know this is confusing, so I'm adding buildbot step text to clarify this:
https://chromium-review.googlesource.com/c/517608/1/cbuildbot/stages/android_stages.py#192

Or if you think we should always run HWTest even if no uprev will happen, please reopen.

To be honest, I think it is confusing. The Android PFQ seems to work differently from the Chrome PFQ. Now this might be a good thing. It appears to me the Android PFQ keeps accumulating OS and Chrome changes without testing them on the actual PFQ boards, but ultimately also without displaying them properly once a hardware test actually runs (the build page will only show the last few OS changes, right)? I don't have strong feelings: not running HWTest has its benefits. But something doesn't seem quite right with the current setup.

Yes, I definitely agree it's confusing... and I guess ExitEarlyException is evil. Maybe it's better if we could mark remaining stages explicitly "skipped", though such cleanup seems to cost more than benefit IMO. Any good idea to clarify this more is welcomed.

I might be misremembering (and have too much on my plate right now to go and veirfy), but I feel like the logic should be the same between Android and Chrome PFQs.  I based the original code on the Chrome PFQ, including the skipping logic.  Could it just be that NOP Chrome PFQ runs just don't happen as often?

NOP Chrome PFQ runs happen after a successful uprev. So I went to the waterfall: they behave exactly like the Android PFQ, no HWTests are run.

So we have consistency. Nothing to see here. WontFix.