Instance unshelve, creation, or migration fails with error "no valid host" whereas compute nodes have sufficient resources to host it

Issue

• After scaling-in and scaling-out a compute node that had its instances evacuated
  • Instance unshelve, create, or migrate fails with error "no valid host" and it's confirmed the resources are sufficient
  • There are orphaned allocations on a resource provider for which the compute can not be found
  • The resource provider deletion was skipped or failed during the node removal