I discovered a possible bug in ROS 6.40.5.
I had to create redundant connection with our ISP, i was advised that we get a public /28 subnet, and we should advertise it on our two router. The main router is a CCR1009, on recent-ish RC ROS, uplink comes in ethernet/sfp como port, i named it ether0, our /28 subnet is bridge-vlan interface (like /int bridge add name=br1 vlan-filtering=yes, /interface bridge vlan add vlan-id=303 tagged=br1,ether2... /interface vlan add interface=br1 name=vlan303 vlan-id=303; /ip addr add interface=vlan303 address=A.B.C.D/28)
Set the peers, AS number, and added this network to /route bgp network (not touching the default syncronize)
It worked fine, ISP got our advertisement.
Then the backup router got us. BGP is configured the same as the main with the obvious differences like router-id; but for testing the backup link i just created a loopback address like /int bridge add name=loopback1; /ip addr add interface=loopback1 address=A.B.C.D+1/28)
ISP didn't get our advertisement, which i confirmed by dumping the BGP communication.
If i know well, the syncronize flag in /route bgp network filters our that networks, where there is no route to in the routing table. In both case, the routing table has the A.B.C.D/28 with DAC flags (Dynamic, Active, Connected) with distance=0, only the interface was different, loopback1 vs vlan303. After setting syncronize=no on that network, the ISP got the advertisement and the router became available.
My MTCINE is next week, so i'm not entirely sure about this, but seems like a bug to me.