The current swarming trigger request often fail with DEADLINE_EXCEEDED:

2017/03/15 13:25:22 INFO: [driver] Trigger request (run ID: 5852562955698176, Worker: Hello_Ubuntu)
2017/03/15 13:25:23 INFO: [driver] Created worker isolate, hash: "2139ca0ef78ffe0ba795ec0f49a5276c490c6a18"
2017/03/15 13:25:23 INFO: [driver] PubSub userdata for trigger: "CICAgICA3LIKEiVodHRwczovL2lzb2xhdGVzZXJ2ZXItZGV2LmFwcHNwb3QuY29tGihlZTIwZDQyOWExMmYyNWRhMGIxY2M5ZTU1NzQ5YmE4ZjMwY2ZmZGY5IgxIZWxsb19VYnVudHUqJmh0dHBzOi8vY2hyb21pdW0tc3dhcm0tZGV2LmFwcHNwb3QuY29t"
2017/03/15 13:25:28 ERROR: failed to trigger swarming task: Post https://chromium-swarm-dev.appspot.com/_ah/api/swarming/v1/tasks/new?alt=json: API error 5 (urlfetch: DEADLINE_EXCEEDED): The read operation timed out :: {"error":"Post https://chromium-swarm-dev.appspot.com/_ah/api/swarming/v1/tasks/new?alt=json: API error 5 (urlfetch: DEADLINE_EXCEEDED): The read operation timed out"}
2017/03/15 13:25:28 ERROR: [driver] Failed to call Driver.Trigger :: {"error":"rpc error: code = Internal desc = failed to trigger worker: failed to call trigger on swarming API: failed to trigger swarming task: Post https://chromium-swarm-dev.appspot.com/_ah/api/swarming/v1/tasks/new?alt=json: API error 5 (urlfetch: DEADLINE_EXCEEDED): The read operation timed out"}
INFO     2017-03-15 13:25:28,933 module.py:806] driver: "POST /driver/internal/trigger HTTP/1.1" 500 -
WARNING  2017-03-15 13:25:28,933 taskqueue_stub.py:1981] Task task2 failed to execute. This task will retry in 0.400 seconds

Failure causes the task in the launch task queue to be retried. The failed trigger request tends to succeed despite the failed request, and the end result is a more than one swarming task.