1. The downside of the "GOP has made a concerted effort to get their votes out early" argument is that it means the GOP has an effective ground game and/or GOTV operation. And I've been hoping that the GOP's ground game was terrible.
That said, there are a few things here. First, in GA, Trump is using Kemp's GOTV and that's a solid operation. In MI, by contrast, the state party has nothing. I don't know how much it has in PA either, given that the party was shellacked in 2022 with horrible, horrible candidates who raised no money. So if GA turnout is good for them, it doesn't mean it will be good in MI and PA.
2. I've seen nothing from the Times to suggest that it's overcompensating for the "silent" voter effect. Other polls are doing more (this is the import of the recalled-vote weighting). It's possible the Times just has a method that will skew a bit right. I mean, it is weighting on a LOT of different categories. I'm not sure how they do that, exactly. The only way I can think it can be done is by disaggregating votes. That is, you model a vote as consisting X% of race, Y% of gender, z% of homeownership, q% of age, etc. ,and then you place weights on each component and then sum them up again. Is that a good method? I have no idea.
Also, keep in mind that the NYT polling averages have Kamala winning in NV, PA, MI and WI. By less than a point, but still.