Also (double-posting here to avoid a big edit), the story points out that the new registrations
last week leave the Ds with a 4% increase over the previous November.
Last week there were about 20,000 new registrants, and 30,000 people switching their registration for a total of 50,000 new Ds.
So 4% increase - maybe 5% with this week's additions - means over 4,000,000 Ds in the state. But the new Ds won't be 5% of the total number of voters.
In 2004, about 800,000 Ds voted in the Primary elections. If there are 100,000 new Ds after this week, that would be 12.5% of the total number of voters in 2004.
Well, in 2004 it was less important as a primary -- and of course, this is assuming that all these new registrants vote. Most of the switchers will; they are highly motivated. Some of the new registrants won't, though. But even if new Ds represent 5-10% of the total, that's still a massive number, in an election! So one possibility is:
The polls might be even less accurate than they usually are.
It's got to be a higher degree of difficulty for the pollsters.