Should it be possible to force delete an instance from a dead cluster node?
When I try (client version 6.17, server 6.15), I get the following error:
incus delete --force app-dev:foo
Error: Failed checking instance exists "app-dev:foo": Missing event connection with target cluster member
Ideally I’d like to (force) remove foo from the cluster, potentially recreate it on some other node, and then when the down node comes back and syncs up, it knows not to start the VM anymore.
No, that’s not possible. If the instance uses no local resources on that machine, you should be able to move it to another server then delete it from there but given your post, that seems unlikely to be the case, so the server needs to be involved to correctly check for any dependency on that instance and to correctly delete it from storage.
If the server itself is dead for good, then you can forcefully remove the entire server from the cluster which will then clear any DB record of anything that server used to run.