> would likely be more amiable to creating shared libraries.

Why's that? There's a gc/no-gc barrier to cross, and also being able to use other features in an implementation doesn't make creating a C interface harder.

I was thinking more along the lines of compiling Tomo code, then being able to link against that pre-compiled binary from other Tomo code. Basically being able to substitute a source file module for a binary module.

I don't know if Tomo supports anything like that, but not having generics would make it easier/simpler to implement (e.g. no need to mangle symbol names). Note "easier/simpler", Nim can also "precompile Nim libraries for Nim code", but the resulting libraries are brittle (API-wise), and only really useful for specific use cases like creating binary Nim plugins to import to another Nim program, where you'll want to split out the standard runtime library [0] and share it across multiple images. It's not useful for e.g. precompiling the standard library to save time.

I know Nim has been working on incremental compilation, too, but I don't know what the state of that is. I think it might have been punted to the rewritten compiler/runtime "Nim 3".

[0] https://nim-lang.org/docs/nimrtl.html