Status update: as expected, my current video encode failed. I'm pretty sure I simply lack the hardware to do a re-encode (and maintain a usable PC for other family members), so I'm looking at solutions that do not involve re-encoding (on the upside, the resulting video quality would be better if I avoided this anyway)
The downside is that I can cut without re-encoding only on i-frames, which means a frame-perfect cut is impossible. The actual, theatrical, cut is here:
"...evil zoot! oh, sh[CUT]e is a bad..."
...and I can get an i-frame cut positioned on either side of that:
"...evil zoot! oh[CUT], she is a bad..." (early)
"...evil zoot! oh, she is a b[CUT]ad..." (late)
The cut at the end is right on an i-frame, so everything will be frame-perfect on that end.
I will keep marching forward with this for my own amusement, but I'm no longer even trying to be perfect so I won't blame anyone for losing interest at this point. I will try to use the "early" cut point because it seems more natural and it also doesn't include any frames from the non-theatrical insertion, so you could convince yourself that a real cut-and-spliced projection print may have looked just like this ;) However, the early cut may not give me as much headroom for splicing in the audio, so I may have to settle for the late cut depending on how that works out.