I think that in between these extremes there is a very wide ground of using specific software tools that are specific to geosciences, which should be mostly on-topic - but with a careful eye towards things drifting to Superuser territory. (and perhaps, as others have identified, a more lenient eye towards Stackoverflow, gis.SE, and things like the R and MATLAB proposals if they ever get off the ground)
I've just asked a question that I think falls somewhere between scenarios 1 & 3 above, which might or might not serve as a useful example :-)