Does team UZR require less of a sample size than an individual player's UZR (3 years)?
I don't really know, but I'd think yes, because in any individual year you've got the aggregate of all the plays the team made. In general, more data = better. So it's not that it requires less of a sample size, it's that it gets to the sample size threshold - the point above which the data start to mean something other than reflecting random outcomes - faster because every play the team makes or doesn't counts into the sample size.
That said, it also pulls in a lot of other random factors. A team's UZR is by definition aggregate, so it counts all the plays made by every player who made a play - that's not a true reflection of a team's "best" defensive lineup. For instance, the Nats team UZR includes all the plays where Lombo played 2nd and Espi SS AND all the games where Lombo played LF. So it works ok as an aggregate reflection of what the Nats have done, but works less well as a projection of what any given lineup will do, since Lombo currently isn't in the lineup at all.
