Failed retrieval of a policy (e.g. after policy change) leads to search index being "emptied out"

[thjaeckle]

We encountered in the following case:

• a policy is used by lots of things
• a change it made to that policy
• as a result a lot of search index updates are triggered for all things which used that policy

If now some search "ThingUpdaters" may not be able to retrieve the policy

• then following WARNings can be seen:

Got AskTimeout during ask for message <SudoRetrievePolicy> and entityId <org.eclipse.ditto:shared-policy / policy> - retrying.. : <Recipient [Actor[akka://ditto-cluster/system/sharding/policyProxy#-1059358312]] had already been terminated. Message of type [org.eclipse.ditto.policies.api.commands.sudo.SudoRetrievePolicy].>

• followed by INFO logs:

Computed - due to missing enforcer - 'emptied out' ThingWriteModel for metadata <Metadata [thingId=org.eclipse.ditto:my-thing, thingRevision=491, thingPolicy=PolicyTag [id=org.eclipse.ditto:shared-policy, revision=4], allReferencedPolicies=[PolicyTag [id=org.eclipse.ditto:shared-policy, revision=4]], modified=null, events=[], timers=[0 timers], ackRecipients=[], invalidateThing=false, invalidatePolicy=true, updateReasons=[UNKNOWN, POLICY_UPDATE]]> and thing <...

• and then the thing in the search index basically is "emptied" - it can no longer be found
• this situation does also not "heal itself" - so the eventual consistency is broken here

This is obviously a bug.
How to fix it however I do not yet know..