This is most likely several different issues, and most of the failures are probably not serious.
500s occur frequently on many task queue paths.

Many of these seem to be "transient" timeouts:

/internal/tq/timers/internal.tq.TimerTask
  e.g.  {"error":"Call error 11: Deadline exceeded (timeout)", "InvID":9093296243831808240, "JobID":"chromium/win-rel"}
/internal/tq/triages/internal.tq.TriageJobStateTask
  e.g. {"error":"Call error 11: Deadline exceeded (timeout)", "JobID":"v8/chromium-trigger"}
/internal/tq/triggers/internal.tq.EnqueueTriggersTask
  e.g. {"error":"Call error 11: Deadline exceeded (timeout)", "JobID":"art/master-gitiles-trigger"}

Others are unexpected responses from dependencies:

/internal/tq/launches/internal.tq.LaunchInvocationTask
  e.g. {"error":"rpc error: code = Internal desc = unexpected HTTP 403 from Gitiles", "InvID":9093296106384992432, "JobID":"quickoffice/master-gitiles-trigger"}
/internal/tq/launches/internal.tq.LaunchInvocationTask
  e.g. {"error":"buildbucket error \"BUILDER_NOT_FOUND\" - Builder u'ios-device-goma-latest-client-clobber' is not found in bucket u'chromium/ci'", "InvID":9093295707406989792, "JobID":"chromium/ios-device-goma-latest-client-clobber"}

And others, too many transactions on same entity group? (Not sure)

/internal/tq/crons/internal.tq.CronTickTask
  e.g. {"error":"datastore: concurrent transaction", "JobID":"turquoise/gi-zircon-roller-trigger"}


Vadim (or anyone), do you know which of these might be important, or could be separated into different bugs?