Is the use of non-paved courses or sections of courses is a real problem for us? I have measured about 10 courses over 20-years that were not entirely on paved roads. I have measured but refused to obtain certification for a few others because they had sections of un-paved road or path with un-defined edges. I have also obtained certification of a couple courses using a steel tape for the unpaved (grassy) sections. These sections were fairly long (>500-ft) and required the use of pins to obtain the shortest route.
The most recent example of a course with un-paved sections is a half marathon with 2 dirt roads totaling over a mile of the course length. These roads are traveled, graded, hard dirt with stone walls or ditches on the edges (well defined edges). I didn't use a different constant for the dirt roads as they are nearly hard as the paved road. The dirt road surface is of different and varying roughness compared to the paved road. This would probably account a difference in constants, but the variation of dirt road roughness would make calibrating on the dirt impractical.
Although Mike Sanford's 3-level idea doesn't address repeatability, it could be a solution if there really is a need for one. I'm sure that a set of RRTC rules for measuring, reporting and checking would need developing for such an approach.