Garbage collection of lost endpoints should be implemented

[Soulou]

Either:

• A specific command to trigger it
• A recurring job that each agent run on its node

Having endpoints existing in database and not on the hosts leads to problem, especially with sand-agent network-connect which selects one endpoint to enter the network. If this one does not exist anymore, it fails with a timeout.

[EtienneM]

@Soulou I don't really understand the description of the issue. What is a "lost endpoint"? Can you provide a way to reproduce the issue?

[EtienneM]

Friendly ping @Soulou on my above message