While technically possible, this operation is not exposed through the CLI (it can be accomplished
with `rabbitmqctl eval` and some function calls). However, the way you framed the question suggests an assumption, which is not correct:
How connections are distributed does not depend on how the leaders are distributed[1]. If you indeed have the leaders of the busiest queues
on a single node, rebalancing them could be beneficial, but it's more complex than that - eg. if we spread the leaders more evenly,
but some publishers are currently co-located with the leaders, they wouldn't be after the rebalancing, so the overall
effect could be negative.
[1] for completeness: Stream protocol has a built-in leader discovery mechanism and publishers automatically connect to the stream leader node,
while consumers try to connect to the followers; so stream connection distribution would be impacted by stream leader distribution