From the minimal info so far I am not running to embrace that failure analysis "According to the shaft manufacturer ."
With all due respect to experienced and sincere local shaft manufacturers, I'd be desperately interested in what the joint MANUFACTURER (hopefully an Experienced Name Brand Supplier (ENBS)) has to say after they had inspected some totally failed joints AND a few joint with significant hours on them.
The FIRST 2 steps I would take would be to get the joint manufacturer's ENGINEERING DEPARTMENT involved. First to evaluate your load/angle/speed requirements, and also with failure analysis. I think getting them a joint that has run a while at maximum conditions may be the BEST thing I could possibly do to identify the "problem."
--------------------------------
Is the alignment change in operation 0 - 17 degrees?
Could you change the initial alignment of the components so the running change is +/- 8,5 degrees instead?
What allowances does your shaft or system have for axial motion (plunge)? There may have to be a spline somewhere. Even a little thermal growth can subject the U-joints to TONS of axial force.
---------------------------------
Unless the shaft has been built/rebuilt with a quality name brand U-joint then frankly all aspects of the metallurgy and manufacturing of the U-joint are suspect.
On a regular roller bearing ALL I need is shallow or improper hardening, crappy or damaged surface finish, rollers whose diameters vary by a few microns, or rollers that do not have the proper crown to cut the hoped-for load capacity in heavy duty service to almost nothing.
Again, with no dis-respect to the shaft manufacturer or your technicians, improper assembly can doom a perfectly adequate design.
----------------------
Are YOU satisfied there is really loss of lube? Do you see a lubricant mess surrounding the plane of the joints, indicating lube is being flung out during operation? As I think about a regular joint it looks to me centrixxxal effects and the orientation of the bearing caps would tend to keep lube >> in << the joint AND the seals seated against the caps. If the failure is really simply loss of lube I'd expect a seal change/modification would make miraculous improvements. For instance, if the seals are all rubber and are indeed popping out of their grooves, then a few drops of crazy glue, or epoxy encapsulation, or some external reinforcement would sure help.
If they are re-lubricatable, could they be re-lubed at 1/4 or 1/2 the minimum failure time ? Re-lubrication would help combat lube loss, lube damage, and contamination.
Grease "life" is murdered by high temps and high rpm.
How hot does the joint run? Here is a good use for one of those finicky infrared laser thermometers. (The other is as a cat toy). If its running Much over 120F I'd be moving away from standard conventional greases, and you certainly need grease with a higher base oil viscosity. Again, I believe the joint manufacturer should be consulted.
How many hours to failure? High speeds in ball and roller bearings tear up grease by shearing action. Note the "speed" of the little uncaged rollers in the joint result from effects of the angular offset X the rotating speed. If there were no offset then the rollers would not rotate at all. Some Rotation is considered essential to keep the rollers' contact points lubricated. Again, The joint manufacturer will be able to tell you what conditions lead to excessive roller speed, and what lube is correct for your application.
----------------
Semi-Long shot - you have a torsional vibration problem that is making the peak loads many times higher than expected. One symptom might be edge loading in a well used joint.
