I do hear the concerns about this encouraging developers to release broken games and patch them later, but I wonder if a reviewer should be thinking about those consequences if they're writing for consumers. I don't have an answer to that.
And also, the damage of releasing a broken game is damage
done. The idea that the internet's etch-a-sketch would be totally shaken if ever a reviewer changed a score isn't realistic; there'll still be tons of scores that don't change (be it policy or viability) and there'll even be hard-liners commenting on those changed scores arguing about how they got burnt by the launch and don't feel a game should get a second shot.
A huge quantity of a game's success depends on its launch performance. It's just not realistic to believe that developers are looking at potentially amendable review scores as a tactical method for releasing a game in a shitty state. If a developer is able to make a comeback after a bad launch and they get the credit for the effort via updated opinion writing, then swell, but it's a huge hole that they'd have to dig themselves out of and no guarantee that there's solid ground above to reach.
I also wonder how it would work if reviewers gave games two scores: one for offline players and one for players with online.
It'd be a nice reader service, but the "business" of reviews is so entrenched in old methodology that it'd take something radical to change it. Advertisers and retailers want the simplest method of getting the message to already-fickle customers, MetaCritic wants one big number to splash on the page (I'm not sure why we still care about MC, but it still has some cache,) Google won't be adjusting its scorebox methods to accommodate game sites going off the beaten path, and even readers balk at taking seriously any complex method of scoring (we used to have places like GamePro and IGN/GameSpot doing different value scores, or EGM - still Famitsu - doing multiple reviews in the same magazine; nowadays, readers get mad if one review site's score is wildly different from the consensus ... egads, the Tomatometer is ruined!)
Any site could do split reviews (are there not sites that do?), but chances are the results won't won't be utilized well by sub-services or readers.