When HC gets a timeout for some call from VPP (e.g. missing plugin in VPP). HC initiates it's restart mechanism. However, that tends to crash VPP since there is a nameless connect call in VPP.

JVpp needs to be checked, how disconnect + reconnect is performed.

Observed on Centos7. But can be simulated if only vpp (without vpp-plugins) ins installed and HD has SnatModule enabled