Why is rust's type inference better than swift?

Recently had a conversation comparing swift and rust's type inference & it made me curious about the details.

Rust seems to give up sooner in the face of ambiguity, & swift seems to push forward and attempt to resolve ambiguities instead of returning an error.

I'm curious about the decisions made that make rust's type inference feel more reliable, despite it being turing complete & not guaranteed to finish?