Please link to specific examples. For instance, you seem to be talking about either http://cautotest/afe/#tab_id=view_job&object_id=63212560 and/or http://cautotest/afe/#tab_id=view_job&object_id=63109569

The first of those two seems to be running just fine. http://cautotest/afe/#tab_id=view_job&object_id=63212560 It still has some queued jobs. I notice that in its debug log (https://pantheon.corp.google.com/m/cloudstorage/b/chromeos-autotest-results/o/63212560-ketakid/hostless/debug/autoserv.DEBUG) that it was run in a mode where the suite job doesn't wait on child jobs. Was this intentional?

http://cautotest/afe/#tab_id=view_job&object_id=63109569
(job log: https://pantheon.corp.google.com/m/cloudstorage/b/chromeos-autotest-results/o/63109569-ketakid/hostless/debug/autoserv.DEBUG)
I see that its child jobs were all aborted.
Picking an arbitrary one http://cautotest/afe/#tab_id=view_job&object_id=63109573 seems like it was aborted before ever starting, my guess is due to timeout.

What exact command line was used to launch these suites?

I guess the last (4th... or so) re-run of the suite succeeded. If you look back at the previous runs they were aborting across the board.

This is the script we use to re-run entire suites. For individual tests we use catotest/

for BOARD in $BOARDS ; do /usr/local/google/home/chromeos-re/paygen/buildtools/src/third_party/autotest/files/site_utils/run_suite.py --board $BOARD --build $BOARD-release/R48-7647.70.0  --suite_name bvt-cq --file_bugs True --pool bvt --no_wait True --priority CQ --timeout_mins 600 --retry True --num 6; done

You shouldn't use --no_wait, that causes the suite job to end without waiting for results, which is why the suite job logs in #2 are uninformative. If all you want is for run_suite to return immediately, use -c instead.

With only the suite job logs that were run with "--no_wait" this is basically impossible to diagnose in detail.