(The commit message says "wip", but no more progress is required: this is ready for review and merge. Thank you in advance!)

Hello Falk, the .qmd and the index.md file now contain different code for the wrap_difference() function. [...] Should the two not be the same?

Indeed. Thank you for checking @bserrien , and sorry that I missed that!

I believe the index.md file is correct (%% extent).

Which of the two formulas is correct depends on the usage of extent on the rest of the code (and whether extend is defined as half or full wrapping period). Here, spatial positions can be positive and negative, and thus the full period (analogous: "diameter") is twice the extent, and we shift by 1* extent ("radius").

Just to make sure, I have tried rendering both versions again.

((Y - X + (extent/2)) %% (extent)) - (extent/2) produces the following wraps:

... clearly there is some unwanted periodicity with half the extent.

Conclusion that the formula in .qmd was correct, it must be: ((Y - X + (extent)) %% (2*extent)) - (extent))

Is the space you imagine for this wrapping something like the game of snake?

Nice analogy; and maybe the availability of old nokia phones in my teens has lead to a subconscious "snake" image. But that was not my first thought.