CCR1009 active-backup bonding intermittent connectivity with no load

I have a new CCR1009 that I’ve setup with active-backup bonding with two switches using the SFP+ and SFP ports. The intent is to use 10G as the primary and fail over to 1G if the 10G switch fails or is being updated.

Config (no firewall, nothing else, just this):


/interface ethernet set [ find default-name=combo1 ] combo-mode=sfp mac-address=mac_of_sfp+ (bonding changed this)
/interface ethernet set [ find default-name=sfp-sfpplus1 ] advertise=10M-full,100M-full,1000M-full name=sfpplus1
/interface bonding add mode=active-backup name=bond1 primary=sfpplus1 slaves=sfpplus1,combo1
/ip address add address=ip_addr interface=bond1 network=ip_network

This works as expects, if and only if there is constant load on the network. If there is no [aggressive] load, then simple ping monitoring fails most of the time. ssh and web access timeout or take a long time to establish. If I use ‘/tools bandwidth-test’ from another Mikrotik switch to keep the link busy, then the CCR1009 is very responsive and works as expected. Ping monitoring does not fail. If I disable combo1, then everything behaves as expected.

This CCR1009 is a backup to my main CCR1009. I’m afraid to setup this bond on my primary CCR1009 that is in use for fear that I could have intermittent connectivity.

Have others seen this?

Thanks.

Update.

Downgrading from 6.44.3 to 6.43.16 fixed this problem.

2nd Update. 6.44.x works if all non-primary slaves have neighbor discovery disabled.