I think that in-memory data could be coordinated via Shared Workers. On a server, this is a bit more tricky, since instances are not necessarily on the same machine.