Better way to handle poles?

Nobody has claimed this yet.

Calculating the distance over a pole, e.g., between [0, 89.99] and [180, 89.99] (with 89.99 as the latitude passed into the cheap ruler constructor) gives me a 57% error relative to the Vincenty calculation. (For reference, I used https://github.com/chrisveness/geodesy which seems to be better maintained than the derivative https://github.com/TankofVines/node-vincenty)

I know that this is already documented as a shortcoming, but food for thought: if you can calculate the maximum error that your shortcut formula would give you in advance (which seems likely, given that you have a % error chart in the readme), could you not simply fall back to the Vincenty (or even haversine) formula if the error is too large? That would fix the pole issue... (for comparison, the haversine formula gives only a 0.4% error for those same two coordinates -- and could probably be improved upon by tweaking the earth radius).

Start by reading the documented pole-related shortcoming and the percentage-error chart in the README, then inspect the distance-calculation entry point in cheap-ruler. Compare the cited coordinates against the existing shortcut and Vincenty or haversine results; done means choosing and validating a fallback or other approach that reduces the pole error without losing the library’s performance goal.

Written by the indexing model from the issue text.