Long-lived feature branches are one of the worst ideas ever. Just f... merge it asap, stop developing 3 things in parallel for a month and then try to deploy half of it. Not that I didn't warn them, but here we are. I guess they had to check.
So, some time ago, we decided to replace a random UUID with a hash of something else, so that we can look stuff up by that something. First it was only implemented in one place, then I moved the hashing, so that entire system operates basically only on hashes instead of X.
Thanks to the new branched approach, change in one microservice (of ~35 at this point) was lost and an old version was deployed on prod. Now instead of hash(X), we have hash(hash(X)) all over the place. Good that we still have the original X cause you know, reversing that would be rather expensive in terms of CPU time.