How do you handle tricky FFI memory safety issues in production?

Hi Everyone～ I'm a Rust developer based in East Asia. I've been repeatedly bitten by FFI-related bugs — the kind that only appear at 2 AM in production. As soon as code crosses extern "C", the borrow checker becomes useless, and problems like ownership violations, lifetime mismatches, double-frees, and pointer escapes become extremely difficult to catch. I'm getting tired of debugging these the hard way, so I'm seriously considering building a tool to help solve this problem — specifically something that can analyze LLVM IR to track pointer lifetimes and data flow across language boundaries (Rust ↔ C/C++ ↔ Zig, etc.). Before I invest a lot of time into it, I'd love to hear from the community: * What are the worst or most common FFI / unsafe memory safety bugs you've encountered in real projects? * What tools or techniques are you currently using to catch these kinds of issues? (Miri, cargo-audit, manual review, static analyzers, etc.) * If a new tool were to be built for this, what features or pain points would you actually want it to address? * Any reasons why this kind of LLVM IR level approach might be a bad idea? Any real-world advice, war stories, or brutal feedback would be super helpful. Thank you!