Most cameras read DX codes with conduction sensors (i.e. pins that make contact and carry a small current), not optically. I have one I routinely fool into exposing bulk-loaded Tri-X at EI 1600 using a piece of aluminum foil. If the black coating on the cassette's DX area was thin, worn, or damaged, it's possible the sensor pins might have made contact where they were meant to read "open", causing the camera to misread the code. If you see this happening with a new roll, it's possible (assuming you can rewind the leader without pulling it back inside the cassette, or have a tool to retrieve the leader afterward) to unload and put tape over the black areas of the cassette and get a correct reading; it's also possible to *change* the DX code by scraping off black or putting tape over silver, in case you want/need to change the way your camera exposed a particular film.

Salvaging the ISO 100 film shot at 2500 is problematic -- that's 4 2/3 stops, and I'd be amazed if it was even possible to push APX 100 that far. Best you can do, if it's not possible to reshoot, is develop in the hottest soup you can find for as long as won't cause too much fog, and then try to save the images with high contrast printing or scanning. Yes, it'll be grainy, and there won't be anything in the shadows, and any bright highlights will be completely blocked up. Nothing's likely to change that...

FWIW, I've pushed Tri-X almost this far (TXT 320, accidentally exposed through the base -- developed to EI 5000, to compensate for a real speed when backward in the holder of about EI 16, later reproduced with 35 mm Tri-X 400 as well) using a home-concocted developer made up from HC-110, Dektol, and a couple common household chemicals, but I'd be amazed if you can get APX 100 to push more than about three stops even with this Super Soup.