Determine if second-level accuracy is enough to determine last-write-wins

[joshsmith]

Problem

We introduced #1203 to solve some problems with GitHub's second-level timestamps. We really need to consider whether this is generally enough, or if there are any edge cases that might violate this check.