Most of my recording is done in the field with a remote dedicated unit (AKAI DPS16) I'd love to record in the field at 96 kHz, but because of the reality of disk space considerations, I tend to record at 44.1 or 48 kHz, sometimes at 88.2, as it leaves me with more resources - and it's a live recording. For instance, like many units, mine will record eight channels simultaneously at 24 bit / 48 kHz, but its resources will drop to 4 channels at 96 kHz. In the project studio I usually record at 96 kHz. I always record at 24 bit resolution.
I do most post production work in Adobe Audition 2.0 which uses the highest resolution practicable and is designed to accommodate 32-bit resolution. However if you give it 24-bit to work with, it will save in 24 bit.
The Audition help page has a discussion about bit depth:
Just as sample rate determines frequency range, bit depth determines dynamic range. When a sound wave is sampled, each sample is assigned the amplitude value closest to the original wave’s amplitude. Higher bit depth provides more possible amplitude values, producing greater dynamic range, a lower noise floor, and higher fidelity:
16-bit = CD quality = 65,536 amplitude value = 96 dB dynamic range.
24-bit = DVD quality = 16,777,216 amplitude value = 144 dB dynamic range.
32-bit = best quality = 4,294,967,296 amplitude value = 192 dB dynamic range.
What all this means to me is that my sound will be the best I can make it, if it is recorded, mixed and mastered at the highest bit depth and sampling rates my equipment can deliver. The very last step in that process is the conversion to 16/44.1 for copying to a CD (or worse - .mp3. ugh).
That way I maintain the best possible sound quality throughout, working with the largest data set my gear is able to generate. If I start with 16 bit resolution or with a low sampling rate, then moving it to higher grade somewhere down the line just isn't going to help much, as I'll just get a better picture of a poor sample.