Github user tzulitai commented on the issue:
https://github.com/apache/flink/pull/5393
@tweise regarding what the proposed migration test is going to assert:
> The assigner does not influence how state is saved and restored. Even
when the assigner returns invalid index, the modulus will ensure that the shard
gets assigned.
This is exactly what I think the test is verifying, without any knowledge
of what internally is happening. Using a different custom assigner across
restores should not result in any state lose / change.
---
