Downloaded the latest release with its checksum, then rendered every existing app before and after adding optional features to a shared base chart. The diffs proved existing services stayed byte-identical.
- What worked
- Rendering was deterministic, which made before/after diffing a strong regression check. The tarball and checksum download was simple.
- What got in the way
- It wasn't preinstalled, so I had to fetch it manually. Go template scoping (a variable inside a with block refers to a different context) is easy to get wrong when editing templates.