The only sources of SEGV issues so far were:

• polars bugs
• mis-specifying types on the OCaml side in externs

The second case is a bit scary, and I think a naive code coverage metric to ensure that all functions we expose are tested would ensure that we don't get surprised by anything like this. A lesser worry is if we, for example, say that an argument is an int option instead of an bool option I don't think we'll get a SEGV as long as we just pass a None, since the OCaml memory representation is identical. Perhaps OCaml tests -> Rust-side code coverage would solve this, though I'm not sure whether code-coverage tools support this kind of thing.