It's not clear from the superficial coverage of Koza's proposal how it'd deal with the problem of, say, the leading candidate getting 20% of the vote, 8 other candidates splitting the other 80%. (Plurality wins? Runoff? Instant-runoff/ranked-pref?)

The proposal is interesting for the constitutional and strategic implications.

One fun one: who enforces? Consider, for example, that California joins but then winds up not liking the result -- perhaps even believing there's been fraud in some other states against the California-favored candidate. They refuse to honor the compact, casting electoral votes loyal to the California-favored, but nationally-behind (in disputed counting) candidate. Does Congress or the Supreme Court have any authority to ignore these votes, on the basis they violate a compact? Can they compel votes in accordance with the compact if that's required for the compact to have its intended effect? (IMO, the only basis for allowing a compact in the first place would be ultimate state sovereignty in EC matters; in which case, a severeign right to change their mind at any time could also be inferred.)

I continue to think your past examples of close elections aren't good examples of how close things would become in a national popular vote. Those elections were somewhat close by sheer chance: the candidates weren't even trying specifically for a popular vote win. The vote totals from non-swing states had a large random component, changed by protest votes and rational stay-at-homes. If the aim is a national popular-vote majority, then a razor-thin victory becomes what both sides (and their bankrolling coalitions) will professionally aim for.

The whole country would then act like a swing state, and the national vote differential would look (in magnitude) more like a medium-state's differential -- making the <500K (and maybe <50K) differential far more likely. And then you'd have to look for fraud and misdeeds in every precinct, in every podunk partisan corner and machine-politics urban ward nationwide. RFK Jr. thinks 160K votes were swung via shenanigans in evenly-divided 2004 Ohio under everyone's watchful eye. I think cheating up 500K votes nationally, from deep within their respective strongholds. would be far easier for each party. Can the national media tell the difference between a 'legitimate' Bush-61, Kerry-38 victory in Texas and a 'run-up-the-score-by-500K-votes-with-tricks' 64-35 victory?