This is a much discussed idea actually, which has been talked about on the mailing lists several times.
General in-place result editing is practically impossible as far as I can tell. One can do many irreversible computations (ie max) on results, and the input values can be themselves the result of computations, or non-editable values (ie page creation date). It is thus important to clearly identify in which cases such support is feasible and which ones are in scope of the system you will be creating.
An additional complication comes from the fact that the SMW query result and query processing code is not in a very good state. It might be difficult to extend it and use it in other contexts as the current ones, which is what I suspect will be needed (or at least very useful) when doing such a project.
So yes, I agree this would be great to have, and if we can get the functionality for at least simple cases, that'd be fantastic. You are right in that this would be quite some task, which I hope I underlined with this mail.
When moving forward with such a project, I highly recommend
* to communicate with the active SMW devs and other interested parties. Work in a public repo, ask for code review, and design review.
* to work in incremental steps. Getting a small amount of solid functionality in a release that can then be expanded is much better then writing a pile of half-working things that never reach usable state.
* write tests
All these steps are essential in making such a project succeed.