I think it's something we ought to explore, if only for the sake of ahead-of-time scanning build systems.
I have gotten to the conclusion that it's actually a good thing that importable headers are indistinguishable from the source code alone.
Essentially, from the build system perspective, an importable header needs all the same semantics that named modules do, including the need for additional metadata (such as preprocessor arguments) as well as the ability to reuse BMI files.
For that reason, I think importable modules will always need to be explicitly identified by the build system either when they're in the same project, but even more when importable headers are shared with a pre-built library.
IOW, just knowing that a header is importable is not sufficient, the same way that just knowing that a module interface exists is not sufficient.
Daniel