If you're working with historical time-series data you could easily be doing millions of time comparisons. so it's almost inevitable that people are going to hit edge cases like this.