I found some answers and thought I would post in case someone else may benefit in the future.
As it turns out, this feature works exactly as expected on McBSP4. I haven't seen it work on McBSP1 but I suspect that has to do with the odd manner of our 4-wire setup on McBSP1 (because of pin conflicts on the board).
The issue was that I was specifying the pins as OMAP_PIN_OUTPUT (Linux board file). This seems like it should be the correct configuration but in fact it is not. Why is it not actually an output? Well, it is, but apparently it also must be an input. My guess is the SRG drives the signal to the pin (as I was actually seeing clock and frame, just not gated clock) and that signal comes back into the McBSP from the pin (or thereabout) through an input buffer. So if you specify as output, the signals from the SRG get to the pins but don't actually make it back to the McBSP (you won't actually transmit/receive any data either). The key is that specifying the pin as an input makes it bi-directional. This is far from intuitive and I suspect many have had their own issues with this nomenclature. However, it is what it is. The moral of the story is that for the McBSP clock and frame lines, the external pins must be defined as inputs (bi-directional), not outputs, whether master or slave.
Hope this helps someone out there.