This effect -- formation of dyes on oxidation -- occurs with at least p-aminophenol and metol; that's what you're seeing when old developers turn brown with time. The dyes that form are slightly weak in cyan/blue, so the overall color is yellow/brown. I'm rather surprised you can see the colors in unbleached C-41 film, however; I'm pretty sure the dye couplers in the emulsion aren't correct for the dyes formed by developers outside the PPD family. Might be that the dyes are weakly coupling and the bleach breaks them loose from the couplers, allowing them to wash out. You might (if/when you get back to it) explore using a formaldehyde based stabilizer before bleaching. This used to be a required step (though, with correct dye couplers, applied after bleach and fix steps, as part of the final rinse) for C-22, and C-41 until around 2000.
Of course, it might also be that the dyes are coupling to the developed silver some way, which gets removed during bleach and fix -- but that wouldn't explain getting roughly correct-looking colors (the dyes should be random in that case, producing a brown or black image).
Obviously, this isn't going to replace proper CD-4 based color developer, but it's a cool and interesting "this gave an image" sort of finding...