The same problem exists on the RUTM50 on the latest firmware.
I have actually fixed the issue with a local patch of a script used to request DHCP leases from the network carrier. The problem is that a DHCP lease length is not specified, and it appears as though most carriers use a default of 2 hours if the client doesn’t request one. With a lease time of 2 hours (7200s), the network carrier terminates the lease immediately at the expiration time and kills the connection on the backend. In order to communicate, a new lease must be established, and the DHCP client on the router handles this by cycling the interface to establish a new lease.
To fix the issue, the user should be allow to specify a longer lease time that would be permitted by their network carrier. In my case I am now using 604800s (7 days). I have proven already that multiple system are able to stay online longer than 2 hours today.
The fix is to apply a patch to /lib/netifd/proto/dhcp.sh to specify the “-x lease:604800” option (replace 604800 with a configurable time in a formal fix please).