Try the following:
- Record at the highest sample rate possible
- Use SoX to resample the recording file with high quality settings, then do the following on two files separately
- Select the same region in both files as your noise profile
- Instead of reducing the noise at first, leave the residue and export
- In a new project, subtract one residue from another by doing phase invert. You should hear the difference in 20-20000 range
Feel free to do the same but with normal noise reduction. However, resampling also leaves artefacts in the hearing range, so you should also get a snapshot of those and subtract it from the signal you get
As for time stretching, it would take too long to explain, but a practical application of high sample rate recording was done for How To Train Your Dragon franchise. They recorded animal noises and slowed them down significantly (4-10 times). With low sample rate recording (below 88.2) it wouldn’t have sounded nice in the finished product