OK, here's the results of some tests and measurements over the weekend.
Loads that functioned fine:
3.8gr Titegroup, 124gr Berry's PRN: OAL 1.161"
4.2gr HP-38, 124gr Berry's PRN: OAL 1.155"
Fiocchi Extrema, 124gr Hornady XTP: OAL 1.093"
The Bad Load:
3.6gr Titegroup, 124gr Berry's PFN: OAL 1.129"
The transition from the conical point to the shank of the bullet- the "shoulder", so to speak- was ~0.03" further forward on the bad handload than the Fiocchi factory load with a similar bullet shape. (max was correct; the PFN's are slightly shorter than the XTP's, so it's not a direct apples-to-apples comparison.)
Second test: I loaded up some dummy rounds using a few spare Berry's PFN's at different OAL's. I seated one at ~1.13", like the bad handload, and the slide stuck shut!
Upon getting it open- after some grunting and cursing- the bullet was found to have distinct rifling marks on the shoulder. Another test round seated at ~1.12" had slight rifling contact but the slide operated normally. A third test round ~1.10" had no rifling contact.
Conclusion: Venom's rifling theory was correct; OAL was the problem. Berry's PFN rounds need to have an OAL around 1.09"-1.10" for reliable functioning in my M3904.
Hodgdon's online load data calls for an OAL of 1.090" for the slightly longer 124gr Hornady XTP, and since Berry's advises handloaders to use lower-end jacketed bullet data, I'm going to run with this OAL recommendation.