Can you compare the generated files (between your archive and what generated)? Also, are you building in debug mode? If so, the debug information could have references to the project paths, or time-stamps)
I've built the code for the Release Configuration and the Debug Configuration with the same results. I would be shocked if the hex files (i.e. any linked library or object files) would have date/time stamps or project path info in them, but who knows.
Given the above the test condition in my first post, I've done some further investigation and it seems as soon as one deletes the "generated source" path and rebuilds, the .cyprj file changes dramatically.
As per Hendrik's suggestion, please do a diff of the Generated_Source directories corresponding to the two different hex files. That might help in debugging the issue.
Just did a comparison of the Generated Sources directory (with WinMerge) and all files are identical. It looks like the link order of the object files is changing and thus changing the hex file contents. I'm guessing the link order may be affected by changes to the .cyprj that is occurring. BTW: either hex file that is programmed on the target hardware works properly, but as noted previously, it is imperative that I can rebuild identical hex files for the FDA.
I actually created a new project from scratch and performed the above test case and could NOT reproduce the failure. The Intel hex files were identical. Not sure if this has anything to do with the issue, but the project in question was upgraded several months ago from 4.0 to 4.1