Let's cross that bridge when we get there. For one thing, maybe they don't take much care about it because it doesn't really matter. If NY was gonna decide the presidency, then they'd probably actually try to secure the vote.
Of course, as linked, similar errors occur elsewhere. But New York seems to be particularly sloppy, they twerked their 2012 results until the following June.
Having 51 state+DC elections confines state-level problems to one state, and not the whole nation. But this in effect moves the problem of counting around from "national popular vote" is close to "pivotal state is close", as we saw in 2000--nobody really knows the actual vote count in Florida, even after a month of scrutiny.
All in all, it seems like NPV will require more national-level regulation of elections (as Lazar pointed out for other reasons), which may have good externalities.
Edit: With all of the caveats above, in the certified results Clinton has won 65,844,610 votes as compared to Trump's 62,979,636. (This is subject to New York's certification twerking; NC also has a few more votes on their certificate than on their "official" online results, and Indiana seems to have recently corrected an error that was on their certificate.)
The margin is very close to 2.1%, but the precise value depends on how you count write-ins that are reported by county boards but not in the state results. Including them will depress the value very slightly.