On Gentoo, where compiling occurs nearly every time a program is installed, one common method used to speed up compiling is to use the -pipe option to gcc as well as to mount memory onto /var/tmp/portate (the location where files are written before actually being installed). These two measures alone can speed up compile times dramatically, as the disk is never touched in the middle of compiles. Instead, the only time reads occur is as the source is read in, and writes only occur when the final binaries, libraries, and configuration files are installed to their homes on the file system.

If your source files are located contiguously on the hard drive [1] and writes are cached to occur in big blocks, while also using some compile-in-memory techniques like those outlined above, I imagine compile times would not vary much on flash-disk MBAs as compared to those with conventional laptop drives.

[1] This could be accomplished by keeping your project in a compressed file and loopback-mounting it to your project's working directory.