Clients can potentially hang onto an RC for a long time, which may no longer be valid, but the client currently has no way to learn about that expiry. This causes problems for hops=1 mode because you might end up choosing the RC of an expired node.
Related to this: if a lokinet relay is not currently active, it should either let clients know this or else just drop all connections so that the client doesn't use them.
Clients can potentially hang onto an RC for a long time, which may no longer be valid, but the client currently has no way to learn about that expiry. This causes problems for hops=1 mode because you might end up choosing the RC of an expired node.
Related to this: if a lokinet relay is not currently active, it should either let clients know this or else just drop all connections so that the client doesn't use them.