Starting with R2025a you can use the new "any-destination" PollableDataQueue to solve several of your problems. Unlike the old DataQueue/PollableDataQueue the any-destination PollableDataQueue can be created on any client/worker, and any client/worker with a copy can send or receive. You can create a any-destination PollableDataQueue like this:
queue = parallel.pool.PollableDataQueue(Destination="any");
The documentation is here: PollableDataQueue - Send and poll data between client and workers - MATLAB
With the any-destination queue you can more easily limit the number of elements of on the queue. QueueLength is essentially the number of elements that can be polled of the queue on this client/worker. For the old queue only one client/worker can receive, and the value if therefore allways 0 everywhere else. For the new any-destination queue, any client/worker can be a destination, and QueueLength is therefore valid everywhere. You can therefore check it before adding elements to the queue to limit the number of elements on the queue and overwhelming the receiver.
Alternatively, you can create a "control queue" which you can use to send messages to the workers to tell them to start/stop/pause etc. Previously you had to create the queue on the worker for this to work. With the any-destination queue, you can create it on the client, send it to a worker, and use it to send from the client to the worker (or from the worker back to the client).
Finally, the any-destination PollableDataQueue makes it easier to send from one worker to a different worker. (This could also be done with the old queue, but was very cumersome, since you still needed to create the queue at it's final destination, and then get it to the sending worker). Since you can send from worker to worker, you can have one worker producing the data, and multiple workers processing the data to keep up with the producer. You can see an example of how this is done here: Perform Data Acquisition and Processing on Pool Workers - MATLAB & Simulink
