Also my reading now is that "bidirectional" actually is in contrast to Hindley-Milner, which is unidirectional in the synthesis / inference direction

There is the THIRD choice of Go / Zig / Java, which are unidirectional but CHECKING-oriented - the opposite of Hindley-Milner

---

That is, I don't think it makes sense to define Go as "bidirectional" because it does a little inference around the LHS and RHS.

This seems to be another case where two different camps use the same word in different ways

(which is not surprising if you observe that a similar confusion happens with the word "type" itself - https://kar.kent.ac.uk/69703/ )