I have heard junkie soundfonts and excellent soundfonts. I am currently of the opinion that the format is not the problem but rather the techniques used to create the samples contained within the .sf2 format.

As far as working in the sampler's native format, so far I'm not sure I like that idea either. Thus far my experience in dealing with electronic music instruments I find that they have a limited life-span (not in terms of their continued usefulness to a musician, but in terms of the length of time a manufacture produces and supports the instrument). I believe this to be true with both software and hardware instruments.

Given the short life span of sampling instruments, I am more inclined to keep a library in raw wave format with a description of how settings might be applied in a generic sense to a waveform loaded into an instrument. This way my library is using a file format that is not likely do die anytime soon, is broadly supported by nearly all instrument makers, and is easily converted to new or old native formats.

The downside is that my library will require considerable hand assembly effort when I want to bring something into a different native format. But then, the world is already like that now (ie DS864 has a lack of good samples in its native format, same with VSampler).

sluggo do you pitch up or pitch down when you notice the artifacts.

I pitched down. For the most part I'm not too worried about it. I've seen the results from an aliasing test of various samplers, and the DS864 did not fare too well. If I was mixing a track and was looking for a polished final product, I would get concerned about minor audio artifacts. For farting around, I don't really care.