I remember the rolling lock problems. There were some problems early on due to spring failures, this was traced to a batch of bad springs. Since the rolling lock was never implemented with two springs (ala the axis lock's two-spring failsafe), when those bad springs started breaking, the locks were hosed. Later on, people were just seeing execution problems -- locks disengaging due to spine pressure or light spine whacks. It felt like just an overall QC problem to me, the lock didn't seem any fundamentally less sound than the axis and those things were rock solid. I'd bet a rolling lock done by Spyderco would have been amazing, especially if there's a way to re-engineer it to be ambidextrous and have that second failsafe spring. Benchmade has not tried to put a second spring in, and I don't now if there are any design issues to do so.
I don't see any theoretical difference in strength between the axis and rolling locks (or ball lock, for that matter, they all use the same basic principle), other than that the rolling lock's lock bar had a cutout so would be significantly weaker than the axis lock's bar or ball lock's ball. That said, it's hard to believe that the locking bar would give before the pivot, liners, or something else. Reliability-wise, again it seems like theoretically they should all be similar, other than the axis and ball lock's double-springs (and more reliable coil springs in the case of the ball lock) versus rolling lock single springs. Other than that, I have no understanding as to why REKAT was having some trouble with the rolling lock's reliability.