Unsynchronized queues impossible to remove

[cpuydebois]

Hello,

We have a 3-node cluster in version 3.13.4 with around 3,500 queues (Mirrored Classic Queues).

Among these queues, a handful are marked as being out of sync with one of the mirrors.
It is impossible to resynchronize or even delete the affected queues, either via the web console or the CLI.

Details of a queue :

With CLI, the command rabbitmqctl cancel_sync_queue <my queue> failed with this message :

Error:
{:absent, {:amqqueue, {:resource, "/", :queue, "my queue"}, true, false, :none, [], #PID<13176.6615.0>, [#PID<13612.8978.0>], [], [:rabbit@alplyolindevb66], [vhost: "/", name: "ha-all", pattern: "^.*$", "apply-to": "queues", definition: [{"ha-mode", "all"}, {"ha-sync-batch-size", 1000}, {"ha-sync-mode", "automatic"}], priority: 1], :undefined, [{#PID<13612.8979.0>, #PID<13612.8978.0>}], [], :live, 0, [], "/", %{user: "Rabbit_user"}, :rabbit_classic_queue, %{}}, :timeout}

and the command rabbitmqctl delete_queue <my queue> never finishes ...

Consumers, who usually create their queues automatically at start-up, systematically fall into error here:

reply-code=404, reply-text=NOT_FOUND - failed to perform operation on queue 'bi-statistiques-parcours-collectif' in vhost '/' due to timeout

How can we restore the situation with these queues that seem corrupted?
We can't migrate to quorum queues at the moment...

Thanks