That number (47.952048 kHz) is the effective sample rate if you have exactly 1600 audio samples per video frame (48.000 kHz at 30fps) and then switch to 29.97 fps but keep the same number (1600) of audio samples. I believe several consumer DV cameras do actually do somehing like this. The XL1s, in particular, was rumored to have some strange effective audio sampling rate. Calling this "drop frame" is misleading because 1) it is confused with drop frame timecode and 2) no frames are dropped - the frame rate is .1% lower.

In the XL-1H I'm relying on the manual which says that in all HD modes the sampling rate is 48 kHz. In the SD modes you have a choise of "locked" and "unlocked" audio but the manual doesn't say what that means beyond that they are, respectively, synchronous and asynchronous. Could be 47.952048... in locked mode (synchronous with the frame rate) and 48.000000 kHz in unlocked (not synchronous) but I really have no idea.

I suppose you could record some of the 1 kHz tone from the camera's generator, transfer to an analysis program and count the samples per cycle. If there are 48 in the same place in every cycle then that's 48 kHz sampling. If the samples appear in different places in different cycles then the sampling rate is other than 48 k.

That's a great idea! Well, after all of this, I've decided it's all not worth the hassle, and there are other factors as well, but for me it'll be the HVX200 workflow. :)