In hockeypuck/hockeypuck#220 I have several users complaining about armored output with randomly-sorted headers breaking workflows that try to detect changes. While it is not recommended to use a bitwise comparison of armored output for this, some environments have tooling restrictions that make it difficult to include an OpenPGP parser into the workflow.

It is relatively simple (and harmless!) to sort the armor headers in the encoder, and it would eliminate the vast majority of such false positives.