The sample editing process was finally completed after many hours and days of intricate work.
Final count - 381 individual samples (gleaned from just under an hour of raw studio recordings). And that's just on the first pass. If I went over those recordings again, I could easily double that number. And it's interesting to consider what would happen if someone else edited samples out of those same recordings, how their interpretation would differ.
Anyway, 381 samples is huge. That could keep me going for years. But in the interests of diversity, as I continue to develop this project in the future I aim to expand the array of instruments and personalities used.
My consideration now is categorizing the samples for quick and easy discovery before and during performance. For now I just use the Mac Finder - where my samples are arranged by instrument type, but it is not at all ideal.
I have heard rumour of someone developing a visual system where samples can appear as nodes that float around, and can be clumped together based on meta-data or audio analysis -- for example instrument type, melodic/rhythmic, sample length, loudness, timbral characteristics etc. So you select one of these elements from a menu, and a cluster of relevant samples forms. Ideally then, I'd just drag and drop these sample-nodes into my instrument and play.