A preset change versus a scene/channel change appears somewhat analogous/ambiguous to the musician using the Axe FX.
Internally to the Axe unit there is a significant difference. When you change a preset, that preset has to be read from storage then loaded into executable memory. Once loaded the preset must be “configured” then the Axe can start using the preset.
With a channel or scene change, the preset is already loaded into executable memory so all the Axe has to do is “re-configure” the currently loaded preset then the Axe can use the new sound.
One consideration in this difference is efficiency. It makes sense to me that the lower efficiency of a preset change would play some role in the small sound gap when doing preset changes. The higher efficiency of scene/channel switching it seems would reduce that small gap... possibly to the point of making the transition imperceptible.
Another consideration is device management. Using presets only, the “textures” (lead, rhythm, clean, punch, etc) for a single song/tone reside across multiple stand alone presets. There is no context that explicitly conveys which presets go together for song/tone X. That context is in the mind of the person setting up the Axe FX. This creates greater reusability of textures, but adds considerable complexity. Organizationally this may make sense to you and there is nothing wrong with that.
Looking at it from the scene/channel perspective, a single preset represents a single song or tone with the scenes and channels providing the different “textures” (lead, rhythm, clean, punch, etc) within that preset. It adds explicit context and relationship to the data. It is a way of conveying to anyone who gets your preset(s) “these things belong together”. This coupled with a meaningful naming convention for your presets and scenes IMO, enhances the overall portability and usability of presets. It also makes your preset library less complex by making the purpose/role of each preset more clearly defined IMO. This approach reduces complexity, but reduces texture reusability.
A final consideration is navigability. The idea of changing to a preset at the start of a composition then navigating scenes while performing the composition is IMO a very natural way to navigate your tones and textures. With a device like the RJM MM GT, selecting a preset can automatically take you to a secondary “button page” explicitly for scene navigation and stomp box access. For me personally the navigability is the real win with scenes and channels. When coupled with a great MIDI controller it can simplify the navigation, thus reducing the decisions that you have to make and accurately execute on stage. This leads to a lower likelihood of making a mistake.
Live performance has always struck me as an exercise in controlled chaos. There are so many variables that are out of your control. Any place that complexity is reduced, I believe reduces risk during a performance. I believe that is a primary consideration.