Comment 20 for bug 1605089

@Kevin, think about this: 2 workers (w1 and w2) are checking neutron DB for same object after applying changes to OVN in postcommit. And at that moment there are 2 new updates happened: the first update change value to v1 before any of them recheck, and the second update change value to v2 after w1 recheck but before w2 recheck. So w1 will see v1 and w2 will see v2. Since there is no lock, w2 can apply change to OVN before w1, and w1 will overwrite it from v2 to v1. This is exactly same race problem. It won't be solved unless keep retrying to confirm there is no more change made to the object after its re-applying.

I agree with you about idempotency.