The cost would certainly be high for a ground-up redesign, but the benefits would be considerable as well.
There is a phase in which to involve photographers in the process. Best is to lay the architectural groundwork that would support most any consideration that a photographer (or designer artist, or, or ...) would want. The usability part of the problem, believe it or not, isn't the hard part.
Besides, if you ask users prematurely what they want, what you end up with is something like "give me what I already know PLUS-or-MINUS something" which leads to an ad hoc death spiral.
Let me give you an example of how it might work.
- Develop an architecture based upon a virtual machine (VM) with a nodal, dataflow architecture.
- The (considerable) features and complexities of the VM can be /hidden or revealed/ in application layers as desired for different levels of users and suited for different purposes.
- The VM allows for a compatibility layer with the historical application where needed.
- The VM allows one to retain, where desired, all intermediate stages of the processing right down to the individual brush stroke, which may be modified in any way at any time. The intermediate stages may be either baked-in or recomputed at will. Sets of actions may be aggregated.
- The VM allows unlimited re-use of intermediate data in more than one stage of the processing. This also allows for any number of derivative intermediates, which retain their dependencies on the original data in every sense, and may be themselves re-used.
- The VM provides all the abstractions needed for efficient concurrent processing without need for special cases.
- The VM allows for a very high-level, object-based scripting language, which in turns allows for sophisticated actions to be built up out of primitives. Programs written in the scripting language may be shared at will. The clear-cut semantics of the scripting language allow users to communicate in clear and unambiguous language about their techniques.
- The application layers above the VM would allow for all the things that one does or can do now, including an outright compatibility layer, but would also allow for sophisticated new techniques.
- The same model could be used for images and video. Moreover, it could be used for media presentation layers (e.g., broadcast, and interactive media) and knowledge-based systems that are unheard of today.