I am a huge fan of what the USC/LAT Panel is trying to do. The data from asking 3,000 people each week who they will vote (with probabilities!) is going to be amazing. But, if their level was an issue before (i.e., too many Republicans) their trajectory is going to be a problem now (i.e., movement will not be meaningful).
I am actually not concerned about the biggest obvious issue with a panel: people feel inconsistent to change their vote. People do not change their vote between major party candidates very much anyway, what we are looking for is movement in the undecided or third party camp and major party candidates. So, I am not concerned about this.
I am concerned that the 3,000 people who answer the poll are now very different from the general population, because USC/LAT has been following them for months. The debate was aimed at listless, likely partisans for Clinton or Trump, who so far are not that engaged and not sure if they will vote. These people cannot exist in a panel that asks them who they will vote for each week. The debate was aimed at third party supporters who think the major party candidates are indistinguishable. These people cannot exist in a panel that asks them who they will vote for each week.
You get the point. The people who move from here to Election Day are the ones who are really now starting to pay attention, un-engaged voters. The USC/LAT panelist have been paying attention for too long to matter in that regard.