Clang isn’t an optional component in the LLVM project - it’s a primary sub-project with the same status as LLVM core. If we want to follow the example of Clang then we should definitely not treat frontends specially. See the Clang contribution guidelines which set a very high bar for adding extensions exactly because of the maintenance issues identified in other replies: Clang - Get Involved.
However, extensions (particularly language extensions) have long-term maintenance costs for Clang. The benefits of the extension need to be evaluated against these costs.
Raising the LLVM project is fine here, but it argues for exactly the opposite process to what this RFC proposes.
I stand corrected on GCC, I’m not really familiar with it compared to LLVM. I’ll let GCC experts comment further on the OSS practices around contributions there. I’d be especially interested if they have similar alternative processes for ‘optional’ components as is proposed in this RFC.