The author, a JVM developer, initially embraced musl for Rust projects due to its self-contained binaries. However, after learning about musl's suboptimal allocator, they benchmarked it on 4-core EC2 VMs. Results showed musl is significantly slower, even with mimalloc, and not just in high-concurrency scenarios. The allocator isn't the only issue; several memory routines are also slow.

The author concludes that while musl's simplicity is fine for small projects, it's a performance liability for performance-sensitive applications like Bifrost, leading them to remove musl as a prebuilt option.