Easy. In the grant you say "portions of this grant will go towards funding a full-time position for 3 years", along with an estimate of how much that would cost.
So my boss just forgot that or thought it would not be valuable when writing the proposal. Most computer science research would benefit from having a programmer. Usually all code is written by students and researchers and (as expected) it is thrown away afterwards. Research and graduating leaves little time to polish code for publication.
Not all projects are that way. Minix, after all, is several decades old and does not fit the "is thrown away afterwards" category.
More specifically, in its early days it was designed as an educational tool. The source code and accompanying text book taught operating systems theory and practice to a lot of people in the 1990s. That sort of code must be stable.
Another example is LLVM. The center of development is at the University of Illinois. Here's the home page for one of the research programmers at UIUC who works on LLVM development: http://www.bigw.org/~jcriswel/resume.html .
