The remaining system-test failure is a little frustrating.. it is intermittent and I often have to run the system tests several times before getting it

      1) should error out for bad etags

  2 passing (5s)
  1 failing

  1) BigQuery
       BigQuery/Dataset
         should error out for bad etags:
     Uncaught DeprecationWarning: Unhandled promise rejections are deprecated. In the future, promise rejections that are not handled will terminate the Node.js process with a non-zero exit code.
      at emitDeprecationWarning (internal/process/promises.js:95:13)
      at emitWarning (internal/process/promises.js:88:3)
      at emitPromiseRejectionWarnings (internal/process/promises.js:120:9)
      at process._tickCallback (internal/process/next_tick.js:69:34)



(node:49173) UnhandledPromiseRejectionWarning: AssertionError [ERR_ASSERTION]: Input A expected to strictly equal input B:
+ expected - actual

- 400
+ 412

Upon logging the response bodies, I found that the server is returning 2 sets of response bodies for the same request:

body {
 "error": {
  "errors": [
   {
    "domain": "global",
    "reason": "conditionNotMet",
    "message": "Precondition Failed",
    "locationType": "header",
    "location": "If-Match"
   }
  ],
  "code": 412,
  "message": "Precondition Failed"
 }
}

body {
  "error": {
    "code": 400,
    "message": "Precondition check failed.",
    "errors": [
      {
        "message": "Precondition check failed.",
        "domain": "global",
        "reason": "failedPrecondition"
      }
    ],
    "status": "FAILED_PRECONDITION"
  }
}

But the weird thing is both responses give the 412 HTTP status code in the response header. When we parse the body, we override the status code returned to the client via ApiError with the error in the body, and that's why the test is failing.

Anyhow, it's probably unrelated to the test failure fixed by this PR, so let's merge this and I'll file an issue with the team.

Woah, that's wild. Filing a bug with the partner team feels like the right move. In the interim, I'm supportive of a retry or skip until the underlying API issue is resolved.