There are three very valid design issues brought up:
        1) In a prototyping part, it is essential that the part be configurable
            to the production part as closely as possible (including the setting
            of a CP bit), but that part must also be restorable to its original
            unaltered state.

        2) Sharing a single die for both prototype parts and for production
            parts is essential. The cost of the prototype part is then just
            the delta of the packaging cost.

        3) Having a UV erasable CP bit means that copy protection is
            reversable, even if you don't have a 'window' for the UV light
            to pass through. (aren't x-ray lasers wonderful tools?)

MicroChip's solution to these requirements was to put a shield over the
CP bit. This doesn't support 1, it achieves 2 (although I suspec the package
is more expensive than a straight window package), and it doesn't address
number 3. Therefore I conclude that it is a substandard solution.

An alternate solution would be to have two CP bits configured in an 'OR'
configuration (there are several easy geometries to support this). One of
them is a floating transistor gate (ie UV programmable), one of them is
a silicon fuse (ie programmable but not reversable). If you program with
CP1 set you get a part that acts just like the production part but can be
erased (including the CP bit), if you program with CP2 set then the part
becomes permanently copy protected. (Note that if you also disable
programming with this fuse you prevent the 'programming' attack as well.)
[NB: If you don't know what the programming attack is, don't bother asking
 since I won't tell you.]

You solve 1, 2, and 3. But it is a new part so there are issues with upgrading
programmers to know about the new CP bit.

--Chuck