I'm thinking regressions and broken tests. Bun is already known to segfault a lot and their existing tests were rather lackluster, the Rust port being just as unsafe would be the least of their problems.
This assumes that the memory safety bugs in the unsafe Rust port are the same as the Zig codebase. A total rewrite with so little review is virtually guaranteed to introduce many new bugs which very well may be more severe than the old bugs.