If you put like that I would use queues over the other two options you stated.
Queue don't lose data if the consumer is a bit slower than the producer and you don;'t need code to prevent a sample to be consumed twice if it wasn't updated before the consumer is ready to process the next sample. With queues two sequential samples can have the same value, with the other two you could never know if it was the same sample or a new one.
If you don't need to have a sample, but only the most recent for another run of the consumer the loacl or property node would be a better choice since you don't need extra code necessary for the queue (like enqueue and dequeue, etc).
Hope it clarifies a little.
Regards,
André (CLA, CLED)