I've assumed similar things, though I recognize the 'turbulence' argument below. In general, it does seem that as one lane slows down a bunch of people jump to the faster lane, quickly slowing that one down.
Regardless, for me the best strategy is to almost never switch lanes. Even when I do switch, 15 minutes later I see somebody that I had passed earlier zipping by me in another lane. There is no value in switching for me that I can determine, and my commute is much more pleasant when I stop thinking about how to get one car length ahead. Whether the lane switchers improve efficiency or reduce it doesn't change that equation for me.
I said 'almost never'. There are a few places on my commute where free lanes are predictable - everyone is merging right to try to get to get onto some interchange, that so several lanes back up, while the people who are trying to stay on the same road can move left and travel quickly. But even then, what are you optimizing for? Maybe 15seconds? Sure, it is more than that in the moment, as you can sit in the slow lanes for a few minutes, but all you do if you move left is rush ahead to the next back up.
