Underrated language for this space: ATS from the ML family, which has a feeling of what if C met SML/OCaml then graduated.

You get more flexibility for memory safety with linear types over affine types like Rust for preventing double free or use after free-like errors (while be general for any use-X-times problems). Refinement types can enforce bounds. Dependent types + viewtypes can build complex, but zero-cost abstraction for your own code or as wrappers over C libraries to make them safe & pushing checks to the compile phase rather than runtme. On top of that, there are proof-level types/values you can interleave in your code instead of using an auxiliary language like Coq or Agda. Compiling/mapping to C you use a lot of the same tooling of C as well as performance charactistic of C (can opt in/out of GC, unboxed types, can layout the memory, as well as TCO); you also get the stable C ABI over Rust’s general difficult to be used in non-Rust projects.

All this to say you have a language that can operate at the system level, type abstractions that go beyond posterchild Haskell, & a proof language to turn those white papers into proof code right in the project. If it didn’t have a special learning curve, it should be a lot more popular in this space.

If your next machine has a higher pixel density than 1080p, the need for aggressive hinting diminishes as pixels are smaller & needing to extrapolate subpixels accurately is less important (and less taxing to compute). That wouldn’t help you now, but in the future you may want to consider something like 2.8k which isn’t overkill like 4k on a small laptop display at arm’s length.