possible temporary lock-out of connections until fd_server quits

[grandrew]

in tunnel.c we have this to search for the SHM index to attach:

shm_conn_info[i2].usecount > 0 && strcmp(shm_conn_info[i2].devname, dev) == 0

but in case all processes quit, we may end up with usecount == 0 but SHM not cleared, and fd_server not quit. In this case we may not be able to attach again until fd_server quits due to timeout.

---

Also check that no one is updating "alive" timestamp so that fd_server really does quit. I am suspicious about long timeouts at sps