Today we only write local EIDs to HC. Remote EIDs are resolved dynamically by VPP (when we send a cross-node traffic).

The problem is that since remote EIDs are still present in VPP after a cleanup, we cannot write new local EIDs into HC in case when

IP of cached remote EID == new local EID IP.

This may happen when VMs swap compute nodes in the next run. Honeycomb will throw exception saying that data already exist.

If a local EID is removed from a node, should all corresponding remote EIDs get dynamically removed from all other nodes?

If yes, the fix would be needed in VPP.

If not and VPP should dynamically update remote EID's TTL or other attribute to 0, then the fix would be needed in HC.