Back in stoneage, while using Windows 3.11 (I think it was calles Windows for Workgroups or so...) there was a bug in calc: 0.02-0.01 resulted in a flat zero. Perhaps it's time to open an issue.
Somewhat related, HP calculators used floating point BCD internally since the beginning, to avoid this sort of thing. (Custom mostly-4-bit CPU chips with specific hardware for the purpose, etc.)
Somewhat related, HP calculators used floating point BCD internally since the beginning, to avoid this sort of thing. (Custom mostly-4-bit CPU chips with specific hardware for the purpose, etc.)