This failure is expected at < M68.
Because those devices have VP8 HW encoding capability, avtest_label_detect detects hw_acc_vp8_enc.
This test at M67, 66,... is kicked off seeing the label detected by avtest_label_detect at M68 and M69 images. However, M67 and 66 images doesn't actually have VP8 encoding capability, and thus this test fails.

Hmm, shouldn't we have the capability disabled for < M68 then? We use test results for gating beta and stable releases, so having these tests failing might be really confusing for the TPMs doing the releases.

 Issue 856164  has been merged into this issue.

video_HangoutHardwarePerf suffers from the same problem, just the error message is different.

soraka 10575.58.0
http://ubercautotest.corp.google.com/tko/retrieve_logs.cgi?job=/results/210952690-chromeos-test/chromeos2-row3-rack2-host3

sentry 10575.58.0
http://ubercautotest.corp.google.com/tko/retrieve_logs.cgi?job=/results/210967421-chromeos-test/chromeos4-row8-rack9-host1

nautilus 10575.58.0
http://ubercautotest.corp.google.com/tko/retrieve_logs.cgi?job=/results/210974648-chromeos-test/chromeos2-row4-rack10-host12

Fails with:
Error: vea/vda_unittest failed

video_HangoutHardwarePerf tries to run vea unittest and then it crashes because VP8 HW encdoing is not available.

[M69]
VP8 HW encoding is enabled by default. All the VEA tests are using VP8 VEA. No failure happens then.
[M68]
VP8 HW encoding is disabled by default. However, I finished applied to all the VEA related autotest at M68. So all the VEA tests doesn't actually runs and instead returns TEST_NA. No failure happens.
[~M67]
VP8 HW encoding is disabled. There are some VEA tests based on the old capability framework. To make the tests pass, I need to cherry-pick my capability CLs to M67. M67 is really about to promote M67. I think the cost of cherry-picking is over than usefulness of getting green dashboard.

Discussing with tfiga@, I close this issue Won't Fix again.