Yeah, you're right. I changed my approach and it worked. Thanks!

Tried in Elixir with various sub-second intervals (decisecond, centisecond, millisecond) and they all failed at some test or other because of rounding issues (millis approach infact timed out in tests, so not sure about that). And I see others having issues like this for a test that should be straight-forward to calculate; my guess is this needs to be looked at again.