A very satisfying Kata to have solved in the end, but it definitely has a "real world" feel. In the sense that the main difficulty is in, first, understanding and often guessing the requirements, and, second, debugging when you passed all specific tests but hundreds of random tests still fail and you have no guide...

In the end, after some incremental progress in debugging I decided to entirely scrap my bounding system for a simpler one easier to debug and all was solved.

Got basically the same thing. It's good to know heapq but for a simple logical task it's nice to be able to do it with no library, though I admit I was cringing a bit at my own till[till.index(min(till))], an obviously redundant and costly operation.

Test cases marked as True or as False by a few decimals points, depending on whether some multiplication is put first or last. That is a serious bug. Test should not exactly compare floats !