This might lead somewhere. Although we experimented with a lot of things, like setting the bridge to use RTSP, then changed all ports to edge. This stopped the BPDU packets, but somehow traffic not addressed to us started to flow through our bridge. This again made the upstream switch scream in horror and shut down our links. Then we tried a config where we filtered BPDU packets, it didn’t work (packets still flew, this could be our error). At this point we reached the maximal downtime allowed, so finished work on-site. I later tried disabling BPDU and inter-port data flow within our bridge on the uplink ports, all led to a blackened state. Might try again if my time allows…
Yeah. Now I simply use srcnat to steer traffic when forwarding. It’s working as intended, but pretty meh. Balancing and bonding is not required (as in: push more than 1Gb BW through a single link), but would be a lot better as I would end up with a nice single (logical) interface. 10Gb is on the horizon, at least. Not in the immediate future, though. And not with the same ISP.