I don't think what Trende is saying is necessarily bad or wrong or anything. He's just going off the data he has, and without a doubt the data he is using doesn't portray a Democratic wipeout (although like he implied, still a sizable win that could flip the House).
I think the difference between him and some of us is that many of us are looking at where we think the race will be vs where it is right now. Also if you weigh the special election results more heavily, it does give you a reasonable basis for believing that the gcb polls aren't necessarily reflecting what the environment is and/or will be in November. I think there is a pretty reasonable chance that the floor falls out from under Republicans when September/October comes around. I think that would be somewhat in sync with past wave elections, where some of them broke late in the cycle.
There is also the PPP argument (which isn't just theirs, really) that the environment is actually stable and more in Democrats favor and the polls are just bouncing around right now as people play footsie with the idea of supporting someone they won't actually support come election day. I don't know how to prove this though.
Obama had 46% approval in 2010, and we know what happened.
To be fair, Democrats were severely overextended in 2010. Republicans aren't in a similar position, which is why a ~6.8% win for Democrats wouldn't flip over 60 seats.
It does't necessarily require the incumbent president to be super unpopular to generate a backlash. Generally the cutoff I've seen is 50%. If the approvals are under 50%, wild things can happen, and there might not be a significant difference between 39% and 44%.
The Republicans aren't overextended at the federal level, but they certainly are at the state level, even more than the Democrats were in 2010, and that's bad. I suspect that having a lot of incumbent senators from the opposition party and a lot of incumbent governors from the party in control on the same ballot is the worst possible combination for the party in control.
