but one question is, is there any assistive technology out there that is actually making use of this facility yet?
They can't unless the browser offers a relationship between the two to AT. Whether any do at this point, I don't remember. There was a big longdesc debate recently where use of aria-labelledby for things like figcaptions were discussed, but each example had something stinky about it. Everyone seemed to disagree anyway. Right now as an AT user your best bet in knowing what the caption of an image or a figure is, is to manually sniff around there and assume the text with most proximity and contextually seems like it is probably the caption text.
This old link shows there was/is work on getting native roles to figure and friends, and that the Mozillians are busy there. But dunno who else.
Frankly I'm still waiting for browsers to properly support TITLE attributes so that they work with keyboard etc. But a decade has come and gone so I'm not holding my breath.
Only AFTER it becomes a standard will browsers start to consider supporting those tags that they didn't decide to support for experimental pages while the standard is being developed.
Should be, Even after... because browsers support something either
1. when they feel like it
2. they're frightened of not supporting something they feel is important and OHNOES another browser already has support for it.
The above also holds for stuff not in the spec. See -webkit-blah-blah drama and the death of Presto.
Besides, if "the spec" means WHATWG, often whether or not someone supports something in the wild already is the basis for speccing it.