It turns out that with my configuration, the mobile connection is unstable. I think this is due to a connection monitoring bug which I will explain. Given that the connection is unstable, it could be that it just happened to disconnect when I clicked on the widget in the slide-out panel.
My configuration is not really "out of the box". There is a twist: the WAN port is connected to another RUT955 serving DHCP. After experimentation I found it's not the DHCP, per-se, which is causing the problem. It is having any IP address and gateway IP configured for the WAN.
When the mobile interface comes up, the software checks its internet connectivity by pinging various well-known IP addresses like 8.8.8.8. If the WAN port is configured with a gateway IP, the pings go out the WAN interface instead of the mobile interface. This is clearly shown in the logs because they give the source of the response. In my case, the response is from the other RUT955 which I configured with the address, 192.168.11.2. The log messages are:
unhandler[2334]: Try to send pings to dest_address: |8.8.8.8| via ipv4: |redacted-mobile-ip|
unhandler[2334]: ICMP packet received from 192.168.11.2
The log clearly says it is pinging via the mobile IP address, but the pings are going out the WAN interface.
When a correct response is not seen for the ping attempts, the software takes the mobile connection down starting with this log message:
daemon.notice netifd: mob1s1a1 (19483): Stopping network mob1s1a1
It later retries and ends up taking the connection down again. I don't know if this infinite loop ever ends because I have not waited long enough.
If, I disable the WAN interface, or even if I remove its gateway configuration, the pings to public IPs get responses and you see this message:
unhandler[2334]: Server 8.8.8.8 reached
In this case the mobile connection stays up and is stable.