Two or three weeks ago, I extended my experiment with a pair of Murata devices at 40kHz, this time with the matching transmitter/receiver MA40S4S/MA40S4R units, rather than relying on the presumed reciprocal functionality of the MA40S4S for the receiver.
In addition, I used an L293 H bridge driver to produce a 64Vpp square wave signal (~10dB more than the 20Vpp I'd used from my AWG). Adjusting the H bridge power supply voltage, things seemed to be reasonably linear when measuring the Vpp at the receiver, but the driver became pretty warm and the MA40S4S also became noticeably warm. From memory, I measured about 1.5W of power was being supplied from the power supply, so possibly about half that was being radiated as sound judging by the toastiness of the devices.
Anecdotally, my cat (about 12yo) was sitting about six feet away during the tests. He could tell when it was switched on and off as his ears twitched in correlation, but he wasn't otherwise perturbed, which is more than can be said when I play my sax, when he scampers away to the furthest place he can find.
From the technology demonstrations we've seen, it would be reasonable to conclude that the aggregated power is 20 to 30dB above what I was using, and certainly operating close to medium saturation especially with beam forming.