Firestore: 'test_watch_collection' systest flakes

[tseaver]

Similar to #6605, #6921.

From this Kokoro Firestore run:

____________________________ test_watch_collection _____________________________
client = <google.cloud.firestore_v1beta1.client.Client object at 0x7ff27e1c25d0>
cleanup = <built-in method append of list object at 0x7ff27e1ac878>
    def test_watch_collection(client, cleanup):
        db = client
        doc_ref = db.collection(u"users").document(u"alovelace" + unique_resource_id())
        collection_ref = db.collection(u"users")
        # Initial setting
        doc_ref.set({u"first": u"Jane", u"last": u"Doe", u"born": 1900})
        # Setup listener
        def on_snapshot(docs, changes, read_time):
            on_snapshot.called_count += 1
            for doc in [doc for doc in docs if doc.id == doc_ref.id]:
                on_snapshot.born = doc.get("born")
        on_snapshot.called_count = 0
        on_snapshot.born = 0
        collection_ref.on_snapshot(on_snapshot)
        # delay here so initial on_snapshot occurs and isn't combined with set
        sleep(1)
        doc_ref.set({u"first": u"Ada", u"last": u"Lovelace", u"born": 1815})
        for _ in range(10):
            if on_snapshot.born == 1815:
                break
            sleep(1)
        if on_snapshot.born != 1815:
            raise AssertionError(
>               "Expected the last document update to update born: " + str(on_snapshot.born)
            )
E AssertionError: Expected the last document update to update born: 1900

[HemangChothani]

closing the issue as solved in 8638

[tseaver]

PR #8638 did not address this issue: it was related to improving how the Firestore system tests clean up resources.

[tseaver]

Hmm, looks like I spoke too soon: changing to use a unique collection ID was part of #8638, and was likely a fix for this issue.