Yeah, I used to have an Arrow engine on my two place. seemed like at least once a season one of the dogbones in the spraue clutch would flip and lock the clutch up. Wasn't a problem really because it never kept the prop from turning, but was a real chore to get the clutch out with that one bone flipped.
What typically happens - especially in a case like yours where it continues to occur multiple times - is vibration from engine, gearbox and prop all come together to form the perfect wave - literally. Constructive wave formation wears out the sprag dogs and/or the races. The tolerances are just thousandths of an inch, and in due time the dogs roll over. In your case the race becomes dimpled so that the dogs grab that same spot wearing it out faster and they roll over in it.
The early Arrows did not have adequate lubrication and wore sprags out very quickly during shut down. The later Arrows lasted quite a while, especially for those who changed the Dexron II cherry juice often. I spoke with several gyro pilots getting 800-1200 hrs on an Arrow sprag.
A guy is now making, selling, and shipping a completely new, sprag-loaded PSRU - before they have been completely tested.
This is a 3-gear design to place the prop shaft as high up as possible on the upright 4 Yamaha, which is great for most FW tractor installations but is strictly verboten on a pusher gyroplane - but the units have been marketed to gyro builders! WTF?
Every gyro builder worth his salt knows that raising the seats and/or engine raises the COG of the gyro, while lowering it lowers the COG and thus creates a greater HTL. In most cases, gyro designers have determined the limits of HTL for a particular airframe and built within that, so that if one were to lower a 165 lbs engine by 3 inches such as in this case then the COG will go down by 1/2 that, and the HTL thus increases 1-1/2", putting the prop thrust line above the acceptable design limits of HTL for safe climb out and push-over, and zero G events.
Said PSRU builder/seller was properly warned of this danger and chose to ignore it, instead pushing his new gearbox for sale to gyro builders w/o warning that his product increases HTL and thus reduces safety on gyroplane installations. This is not the first time for this fellow to do things that are not on the up-and-up.
The same guy withheld information and personal experience concerning Rotax rubber donut failures, as well as other Rotax C box problems and RK400 clutch failures he knew of in the past on his modified installations, while he continued to promote and sell them for years until someone finally blew the whistle on the laundry list of failures and concerns.
The basic problem with sprags in any new, unknown installation is that often when sprag clutches fail they do so without warning, and completely let go so that all power is uncoupled all at once. The new sprag PSRUs mentioned above have three gears, which increase harmonics dramatically over two-gear PSRU's, with no known precedent except to say that sprag failure at some point in time due to prop and gear vibration is inevitable. But when? You may think that automatic car & truck tranny sprags last forever, but those are soft, fluid-coupled applications, not hard-coupled to a very strongly vibrating prop pushing 650 lbs thrust and a set of three very noisy gears grinding out 150 hp at 300 lb-ft torque.
After 20 hours of running and two hundred starts the builder declared his PSRU ready to fly test. After a very short while he found that it created way too much heat - engine power lost to heat in the gears - and pulled it off for a redesign to hold more oil. That set the test clock back to zero hours.
Units were shipped to customers shortly after very little test time.
They have just been recalled for problems.
A few folks in the gyrocopter community know that Arrow sprag gearboxes could be relied on to last 800 hrs before rebuild. Based on that history, until this builder has reached a minimum 800 Hrs on a single PSRU from start to finish, without a single change being made to the test unit other than oil changes, the consumer is advised that the unit is completely experimental and may fail catastrophically at any time up to the known test duration currently completed.
Gyroplane and gyrocopter builders are advised to use PSRU's which keep the prop thrust line as close as possible to COG, and to stay away from increasing HTL unnecessarily for any reason.
I have never seen a single gyrocopter in my entire life built with an unnecessarily tall PSRU which increases HTL for no reason whatsoever.