Has anybody considered making use of the host OS recording sources "What U Hear" or "Midi Synth" for recording audio when capturing to video?
For example, if the game uses MIDI music, and the user starts recording video, have DOSBox automatically record the soundcard's MIDI synth output and mix it with the emulated digital sound for the final file. Granted this would add a bit to the CPU usage when recording video, but at least you'd keep the users happy and less time spent dealing with their complaints means more time for the devs to write code.