The LabVIEW queue doesn't use a circular buffer if you define the maximum size, nor does it preallocate the memory (according to the documentation). LabVIEW 8.6 does have a lossy queue, which can allow you to use it like a circular buffer if you use the preview primitive.
Generally, when you enqueue an element, dequeuing it does not require any additional memory allocation since you have a 1-to-1 relationship. LabVIEW simply uses a pointer to the original memory location. This assumes that the wire wasn't forked before it was enqueued, because then you might get a copy.
Release Queue destroys the reference wired into it (which takes 4 bytes) and if it's the last reference or if you used the force destroy input, the queue itself will be destroyed as well and its memory should be released. You should usually call one release queue for each obtain, but it shouldn't really matter if you destroy the queue, since all those references will be released. In any case, just make sure you don't obtain references to the queue in a loop.
If you want a circular buffer implementation, you can also look at the example I posted here, which maintains a constant memory footprint.
___________________
Try to take over the world!
