I have always found that for really high quality width you have to have some form of stereo processing engaged in order to go beyond the 100% hard L/R boundary. I have repeatedly ended up with that conclusion. But I also find that this should be done during the mixing phase on one/a few of the mix/master busses using hardware and allow the mastering engineer dedicated access to it during mastering. I tend to not like it on pad type elements, I like to use it on sounds that are very short in duration and rhythmic or that it describes the ambience (such as the mids/highs of the drum room/overhead), not on the hi-hat, but for instance the delay of the mid/high frequencies of some rhythm guitars could be quite nice to get out there to further enhance the power of the groove. Also, this could be automated so that it is added with a particular pattern. Placing the whole drum kit in a stereo processing configuration (minus the kick), I think can work quite nicely to separate the drums from the rest of the mix, separating the drums in the mix is a pretty important mix factor in many popular genres.
But you should not have to do very heavy stereo processing, with the right routing, panning and with the right rms/peak/fundamental on the sounds in the mid and side components, it should naturally just sit pretty well given that the arrangement allows it. You need to study the arrangement also to get an idea on how the separation best can be established.
But naturally you get great stereo qualities with hardware based processing. This is because the sounds maintain their unique features more at higher gains. It is just that when you further enhance it in the hardware domain it can become super rich and sweet.
