You can consider the beta-amyloid fiasco as a prime example of biomed's replication issues. ML/AL replication issues largely stem from the source data used for training not being available.

In my field, of psychotherapy, social and cultural factors change over time making perfect replication impossible. I'm working on a paper now and we are trying to account as much as possible for replication of our results.

Actual medical science, down to biochemistry.

Look up the replication crisis - if not outright fraud - in Alzheimer research, as an outstanding recent example.

Everything. When science communication to the public uses capital-S science as the big group of boffins figuring things out, theres not a lot forgiveness for after the fact subtleties.

I’m in no way qualified to answer, but I recall reading about a slew of publications being retracted across all fields after the results couldn’t be replicated. If I remember right there was a team that has become infamous for simply trying to replicate studies and finding a very large number of them can’t be.

Economics