Hi, Peter. This regards a customer's MST network, operating in a ring topology. Each ring segment iscarried over a metroE point-to-point segment, but without link propagation.
One segment of this network was mostly down, but working intermittently. Each time it would start working, MST converged quickly (fast handshake) and traffic was routed over the bad link. But when the link would quit passing frames again (several seconds later), it seems that reconvergence was slow (max_age timer). In a babbling condition (e.g. 3 seconds up, 27 seconds down, repeat), you would spend more time blocking, than forwarding.
Ideally, the underlying network should propagate link failure, allowing for sub-second reconvergence on link-down; and also allowing interface event dampening, to prevent east/west flapping.
I can also imagine using EEM to watch for MST instance root transition from east to west, which would could trigger a high port cost on the east-facing port.