The buzz word or phrase is "Build to Print". That's where you take an old blue print and exactly reproduce what you see. If you change anything then you have to go through the long quality assurance song-and-dance. Doing that is expensive (that's how Lockheed-Martin makes their margin).

Responding to specific points made by dtb:

>1. Landing accuracy?
>
>How much can we improve landing accuracy without redesigning
>the system? Reducing the size of those ellipses would make a lot
>more interesting targets feasible.
>
>-- Better atmospheric models?
>-- Better surface mapping?
>-- Better in flight tracking?

The landing accuracy is pretty darn good as-is. I don't think it can be significantly improved without a major design change. The Martian atmosphere is difficult to model. I don't see any near term improvements there. We currently have a very good surface model of Mars and it's getting better. The flight tracking from the Deep Space Network is very good. Improving it will be difficult.

>Where does most of the dispersion come from, and how much
>can we reduce it with an unguided entry like MER? Have we
>reached those limits yet?

You've already summed it up: aiming and tracking, atmospheric modelling, errors in the aerodynamic modelling, etc. Most of this is hard to improve.

>2. Minor new science
>
>I know this way lies madness, but maybe a couple of quick
>and easy things to bolt on, maybe. The Planetary Society's
>microphone....

MER's science payload mass budget was pretty close to maxed out. Adding more payload is probably not an option. The microphone was simply a "gee wiz" thing. They'll probably do that once with some vehicle like Phoenix and then the thrill will be gone. I think the "true color" stuff is over blown. It's more interesting to select color responses that tell us something about Martian surface geology. They've done this effectively with MER.

>3. Power
>
>If one could swap out the solar cells for an RTG of equal output
>and equal mass, that would be great and worth investing in some
>redesign

This is not an option. Mind you, I'm very pro-nuclear and strongly favor the use of RTGs in Space. However if you put RTGs on MER then you'd have to redesign the whole vehicle from a clean piece of paper. All the sunk development cost based on MER would be lost. I also suspect that once you included RTGs in the design, the MSL design would morph out of the MER design with all the issues associated with MSL. MER is a clean simple design that has been proven to work beautifully. Let's not argue with success but instead milk MER for all it's worth. AFTER the MER design has been fully exploited THEN bring out the RTGs along with the super big (and expensive) rover designs.

Gary