Many years ago I read about Yamaha having small end bearing cage failures in some of their racing engines.  It was attributed to exceeding a certain RPM... even just for a moment.  If that happened the cage would fail soon.  One fix was using loose rollers with thrust washers which is maddening to assemble.  (Husqvarna used this on some engines). 

Eventually the issue seems to have been solved...  they probably figured out a way to make better cages?

Side skew and edge loading.
Secrets are:
To let the needles ride where the small end runs. Cage capture is less.
Limit end play on the needles in the cage. Cage is tight to the needle ends.
and radius ends on the needles, actually tapered on the best.

This allows the cage to be an assy. jig.
Approaching loose needle(crowded) performance, but not quite.

Loose is bestest for high RPM's.

cliff

forgot the needle count went up or down to avoid resonations

I crowded the rollers on the bigend of my 185/200 honda 4 stroke.
Bigend cages would eventually break & lock the rod up when revving them hard racing in miniature roadracing built as a 125.
I had no good roller assembly so cut the cage ends to use as a thrust or distance piece for the rollers then crowded them in.
Had a gap of less than one roller.
Used to ride it with one hand on the clutch waiting for 'something' to happen but it just carries on no problem & eventually sold it.
Think that motor was in the PE250 frame when sold.
photos.app.goo.gl/CwY6eXp2reEcc1Qd6