Given the current state of the test matrix and the low risk/high reward, I'd say that it's worth it, especially if you can add it in a way that is easily turned off at the flip of a switch/comment/define.
It's handled by a new preprocessor block in boost/type_traits/intrinsics.hpp: there are some new intrinsics added (because VC8 goes further than our previous best performer: MWCW), and quite a few tests have had to be updated (the new intrinsics actually caught quite a few bugs in our current test cases). But basically yes, it can all be disabled by just commenting out a few lines in intrinsics.hpp. The biggest danger is that I've messed up the preprocessor logic somewhere - but if I have it doesn't affect any compiler I've tested with so far.
Where is the instrinsic support documented, BTW?
Do you mean Microsoft's documentation (http://msdn2.microsoft.com/library/ms177194(en-us,vs.80).aspx) or the type_traits docs? If the latter, then it's all handled by comments inside boost/type_traits/intrinsics.hpp, but it probably should be in the "real" docs as well, I also need to mention that VC8 handles various traits correctly (currently only MWCW gets a mention for that), but I won't commit changes to the docs until I'm sure about the code: type_traits changes do have the habit of jumping up and biting you at the slightest opportunity :-( Anyway I'll go ahead with the change, and if it bytes back too badly, I'll just revert the lot. John.