Tony Faulkner had a great idea akin to this subject. This takes a bit of visualization and I hope I make it clear.
1. Using a Paqrat or Apogee PSX100 (or any similar device spreading two channels of 88.2k/24 bits across 8 tracks), record your music.
2. Because the stereo recording is laid across 8 tracks, the arrangement of the recording would be this: Even tho' the entire sampling frequency is 88.2k, the effective sampling rate for each track is 44.1k because of the bit splitting between tracks.
Track 1: Left Channel odd samples (Most Significant 16 bits,)
Track 2: Right Chanel odd samples (Most Significant 16 bits,)
Track 3: Left Channel even samples ""
Track 4: Right Channel even samples ""
Track 5: Left Channel odd samples ("bottome" 8 bits, equalling 24 bits when added to Track 1)
Track 6: Right Channel even samples ("bottom" 8 bits, equalling 24 bits when added to Track 2)
Track 7: Left Channel even samples ("bottom" 8 bits, equalling 24 bits when added to Track 3)
Track 8: Right channel odd samples ("bottom" 8 bits, equalling 24 bits when added to Track 1)
Normally after recording, one would play back thru the Paqrat or PSX100 to restore a stereo stream of 88.2k and 24 bits. Instead:
1. Pan tracks 1 and 3 left, with the faders at zero.
2. Pan tracks 2 and 4 right, with the faders at zero.
3. Pan tracks 5 and 7 left, but lower the fader 96 dB.
4. Pan tracks 6 and 8 right, but lower the fader 96 dB.
Listen to the music. You have effectively reduced the sampling frequency to 44.1k, while retaining all 24 bits. Because you haven't used an antialiasing filter, the sound should be more open than it would be otherwise because there are no phase shifts due to the antialiasing filter.
NOTE: Music with a great deal of high frequency content won't sound good with this method because of all the aliases. But classical and jazz should do well.
Barry