The customer has four active/passive pairs of AMQ 7.10.0. They are arranged in a fully-connected mesh. On one (and only one) of the active brokers, tens of thousands of messages are stuck in one of the store-and-forward queues. Whenever new messages are produced to this broker, they go directly to the store-and-forward queue, and stay there. The other brokers are all running, and all the SnF queues show that there is one consumer on each queue.

All the brokers are identically configured, except for the IP numbers of the various peers. But only one of the brokers has messages stuck on the SnF queue.

We tried setting producer-window-size=-1 on the network connector, as this seems to have helped in other installations. Here, however, it made no difference.