@b0rk one of my colleague at a previous job wrote test with two mode, one took some output like an API output and put in a file, the other checked the output was the same as the file content.
I didn't like it as the test was hard to read and the git commit was massive and I'm pretty sure nobody checked the actual output.
If the goal is to make sure the API output stay the same, comparing checksum make it more explicit IMO we just check nothing changed