From my test this issue only shows in Boron, to reproduce:

1) Start mininet 1 switch:

sudo mn --controller 'remote,ip=192.168.0.1,port=6633' --topo tree,1

2) Change FRM configuration (e.g. enable flow stale entry):

PUT http://192.168.0.1:8181/restconf/config/forwarding-rules-manager-config:forwarding-rules-manager-config
{
"forwarding-rules-manager-config":

}

3) Break switch OF channel with iptables:

sudo iptables -A OUTPUT -p tcp --dport 6633 -j DROP

4) Wait until connection in controller is in FIN_WAIT state:

vagrant@tools-vm:~/controller-boron\> netstat -npa | grep 6633
(Not all processes could be identified, non-owned process info
will not be shown, you would have to be root to see it all.)
tcp6 0 0 :::6633 :::* LISTEN 19605/java
tcp6 0 557 192.168.0.1:6633 192.168.0.24:56521 FIN_WAIT1 -

5) Restore connection:

sudo iptables -D OUTPUT -p tcp --dport 6633 -j DROP

Now you will see switch is stalled in controller and only recover action is to restart controller.

Karaf log shows SLAVE state is not propagated to switch and if I look at the wire I do not see any ROLE request message going to the switch.