i'm very intrigued by some of the responses and theories. thanks for the feedback. i'll provide you some more information here, and more of my own opinion.
the load is the same bullet as mentioned here, at a similar velocity.
http://btammolabs.com/tests/4.htm and there is no published load data, so I won't discuss the details, but when i spoke to the mfg, they were surprised and indicated it was very, very hot. Based on the chrony results, and the amount/type of powder, I'm fairly certain these are at the upper end of the normal range and not the lower end.
the brass was 2x fired RP (now 3x fired) from a lot of 10,000 pieces, of which i have fired a couple thousand a 3rd time with no issues. I'll post pictures later today of other spent brass from the same group (i.e. fired same number of times) so you can see how much deeper the normal primer indents from this rifle are, and how much more rounded the other primers are.
it's not the pic that's making the indents look that way. they look odd in person and not very deep, but much shallower than normal.
also, it's not a quick cycle time. it's a mid-length gas system with 16" barrel.
my personal theory is that the the pressure made the primer very loose in the pocket (it takes almost no force to move in and out, like, you can blow on it to move it) and when it was ejected, the centrifugal force of spinning or when it struck the ground caused it to slide halfway out.
snuffy's low-pressure theory is interesting, but i've also put probably 300-400 subsonic and nearly subsonic rounds using the same lot of brass through this same gun, and that hasn't caused any primers to back out.
but to your point about the shoulder being set back too far, i DID run these cases through my sizer die twice during this load process. once to size them so they would fit in my case trimmer and again after they were trimmed just because i'm loading on a progressive and it's easier to let them feed through the case feeder than to stick them in station 3 by hand. however, between and after the sizings i verified it was set correctly using a case guage.