We debated one for a second at the shop but it doesn't really tell you anymore than a gmeter and driveshaft speed sensor do.
I've used these (strain gauge torque meters) in laboratory dyno applications. The win you get over shaft speed and G meter is that you're measuring the independent variable, and not the dependent variable, so it doesn't matter how much wheelspin you have, how much power you're losing behind the shaft, what the coefficient of friction is today, or if somebody forgot to take the starting battery out of the trunk, or you have a passenger and you didn't when you weighed in (throwing off your weight estimate ruins your instantaneous TQ and HP calc). You have an instantaneous RPM number and an instantaneous TQ number at each data point, and your resolution is probably good down to five shaft revs or so, so you have a really good idea of just what is going on within a couple of milliseconds.
Strain gauges are also really easy to calibrate. You pretty much set a zero value and one span value and you're done. We did ours with a torsion beam torque wrench when shaft was vertical and with a weight and arm when horizontal.
It's a lot like the difference between mixture tuning with a colortune versus a WB EGO. Both will get you useful information, but the WB is usually going to be much more accurate and precise.
If I were setting up something like an offshore boat or a Bonneville car I'd be all over one of these - you generally can't simulate the running environment on a dyno well enough, and having known values about just what torque caused how much wheel spin or just where the engine hit some weird harmonic at a RPM sustained for minutes or hours can be very helpful.