The most obvious use of stereo filtering is to ensure that the stereo image established for a voice before filtering is preserved. If I start with a stereo sample and have only one filter for the voice then the two channels are summer prior to filtering and there’s no way to recreate the original stereo image. You can pan the filter output, but each voice would still be a mono signal at a designated position in the stereo field.
So with the introduction of stereo samples it’s really important to preserve the original content. There are other benefits as well, since a voice is composed of two sampled instruments and two DSP oscillators. Each of these can be independently positioned in the stereo image prior to filtering. Being able to have two retuned oscillators in a voice, one panned hard left and one hard right, creates a wide stereo image within a single voice even without considering samples.
Lastly, it is possible to adjust the cutoff of the left and right filters independently. They’re listed as mod destinations, so any modulation source can adjust them individually. There is also a “stereo split” parameter that can be adjusted using a soft knob when in the filter section. This simultaneously increases the left cutoff while lowering the right or vice-versa.
The one thing it seems you cannot do is apply different resonance settings to the left and right filters. This isn’t a hardware limitation, as they can be addressed independently in 16-voice mode, but presumably for simplicity DSI hasn’t included distinct left/right resonance modulation destinations.