This is @WanzenBug and mine solution to a small challenge in school. The task was to add 0.0001 to 10000.0 (both float) repeated 10M times without arithmetic error. As demonstrated by trivial.c, an ordinary loop will fail because of the absorption effect.

We wrote the fastest program, using a minimum heap and running in just under 3 seconds on my laptop, whereas it can take weeks if you get it wrong.