First of all, I agree - it certainly depends what you're trying to achieve.
Where I disagree is the following argument (again, I'm not trying to start a war ; ) )
...where I'm more interested in the pixels as data, it makes more sense to convert the gridded scan topology directly to a pixel array rather than interpolating values on to a new pixel grid based on polar coordinates.
First thought - The elementary observations of 3D-scanners are: a distance, a direction and a tilt angle. Hence, these magic boxes are as polar as possible!
a) You've mentioned the scan topology - this information doesn't fall from heaven. So the question is, how does this information ends up in the file? The manufacturers did this work for you. Sometimes this information is correct - sometimes you have smaller or larger data gaps in between. If you need the correct positions in the image (or e.g. for 3D-triangulation in 2D space) then you have to go back into the world of the scanner - which is a polar one and not a new description.
b) Initially, there's no interpolation. All you do is to sort them into the grid. So there will be gaps that you can of course interpolate, if you so wish.
c) If you should make it to Intergeo this year I'd love to catch up for some bevies and some exchange of thoughts!
Long story short: my interpretation is the inverse of yours - for visualization purposes I'd naively trust the content of the gridded file. If I need reliable data, I'd generate the topology myself.