test(robot-server): Fix race condition in integration test

[SyntaxColoring]

Overview

This fixes a test that was flaky because of a race condition.

We were waiting for a run to succeed by:

1. Polling until its status indicated "done."
2. Checking that the final status was succeeded.

But we would erroneously proceed from step 1 to step 2 when the run transitioned from running to finishing, and finishing is of course not succeeded, so the test would sometimes fail depending on timing.

Test Plan

Just make sure CI keeps passing.

Risk assessment

No risk.

[CaseyBatten]

Looks good to me, appropriate change to increase the scope of that check. Thinking on potential alternative cases, would it be beneficial to adjust the check to include "stop-requested" as a third valid pre-end state on this check? so this:

if latest_status not in {"running", "stop-requested", "finishing"}:
            return latest_status  # type: ignore[no-any-return]

[SyntaxColoring]

Looks good to me, appropriate change to increase the scope of that check. Thinking on potential alternative cases, would it be beneficial to adjust the check to include "stop-requested" as a third valid pre-end state on this check? so this:

if latest_status not in {"running", "stop-requested", "finishing"}:
            return latest_status  # type: ignore[no-any-return]

That would be more appropriate, yes.

And there's a very similar polling loop in a different test file:

opentrons/robot-server/tests/integration/http_api/runs/test_labware_offsets_on_compatible_modules.py

Lines 35 to 49 in cd064e6

	async def poll_until_run_succeeds(robot_client: RobotClient, run_id: str) -> Any:
	"""Wait until a run completes, and then assert that it succeeded.
	Return the completed run response.
	"""
	completed_run_statuses = {"stopped", "failed", "succeeded"}
	while True:
	run = (await robot_client.get_run(run_id=run_id)).json()
	status = run["data"]["status"]
	if status in completed_run_statuses:
	assert status == "succeeded"
	return run
	else:
	# The run is still ongoing. Wait a beat, then poll again.
	await asyncio.sleep(RUN_POLL_INTERVAL)

That one takes the approach of listing out the "done" cases, instead of the "not done" cases.

This should be a shared function somewhere. I will do that in this PR.

---

Edit: Done.

[SyntaxColoring]

Got a 👍 from @CaseyBatten in person.