TL;DR: A poll in Minnesota of 635 likely voters showed her lead after the DNC went from 10 to 5 percent. And it also has a margin of error of +-4.5 percent. So in other words, this is a tiny ass poll with a margin of error almost the exact same size as the supposed lead being halved. I'm not saying it's not possible, but with such a small poll, I don't know how accurate this.
If I understand margin of error correctly, then what this means is that the “true” lead that harris had over trump at the time of this poll has a 95% chance of being somewhere between +(5+4.5) and +(5-4.5).
So according to just this poll, there’s a 95% that Harris’ lead in Minnesota is between 9.5% and 0.5%, so somewhere between “what the polling was nearly already saying” and “almost neck and neck”.
And that’s not even taking any of the other polling into consideration…
Considering there are polls that have 2000 or more people and those are generally considered more reliable, 635 is pretty small. It might not be small in a gym, but it's smaller than my graduating high school class. So it's less reliable as such.