That's another great point. With a large build in tmpfs, you can either run into the size limit of the tmpfs, breaking the build and wasting time, or run out of

That's another great point. With a large build in tmpfs, you can either run into the size limit of the tmpfs, breaking the build and wasting time, or run out of memory and end up with an unresponsive machine. Entrusting this to the VFS cache has better predictability, and nearly the same performance in many cases.