In reading the thesis it appears that considerable potential does exist; I didn't realize that this was called a Cardan mechanism, but knew that considerable work had been done in this arena. Becasue of the necessity of the rod to pass through the crank center, the usual arrangement is of a single overhung gear/crank.
There are of course more parts involved; that may not be a detriment if the aggregate mass and expense of those many parts is comparable to the fewer but more massive and expensive components. A multitude of rotating parts may also be a good trade off if the mass of the reciprocating parts may be reduced and balancing shafts eliminated.
As discussed in the thesis, overall reliability and smoothness will be dependent upon the accurate fit and minimal clearance of the gearsets. In this case, where the gearsets are not crank-mounted, but exist as free-floating assemblies, it may be possible to mount the locator gears on eccentrics so that clearance may be adjusted without the use of high-precision gears.
I also don't see this mechanism as terribly fragile; yes absorbing piston pressure through the medium of a gear leaves me uncomfortable; but, because every individual crank is double-ended, and because those crank gears each engage at two (or more) points it would seem that a number of load paths exist to spread out those impulses, and that engineering in reliability shouldn't be insurmountable.
I make no claims that this approach is best or even better; I suggest only that it seems worthy of some thought. I brought this forward simply because progress seems require re-examintion of our basic assumptions periodically.
