I find this lag is increased on Haswell based machines. From what I can tell, it is an issue with there being a lag time between the GPU going from a low power to higher power state.
When you are just using software normally there really isn't much strain on the GPU, so the OS will clock it down to reduce power consumption and heat. Then when you try to re-size a window or open mission control etc it will lag for a second while the chip is still in this low power state, as the animation wont be delayed until the chip clocks back up.
Once the OS realizes it needs more power, the chip will clock up and the lag will mostly go away, until you allow the system to settle again. I see this all the time when scrolling image heavy web pages or switching from space to space. Laggy at first, then nice and smooth until I let the system idle for a second on a static scene.
Of course, the higher you set your resolution, the more strain will be put on the GPU which will of course increase the lag.
The issue is exacerbated by retina screens as any resolution beyond "best for retina" will start rendering your screen at significantly higher resolutions than the screen itself. For example a HiDPI "1920x1200" setting on a 15" rMBP will render the whole screen at 3840x2400 and scale the resulting image down to your native resolution. This is almost 2x the resolution over a HiDPI "best for retina" 2880x1800 (1440x900 equivalent working space).
5.1 Million pixels vs 9.2 million.
One thing you can do is use applets like "RDM" to set your resoltion to a non HiDPI one. If you want the working space of 1680x1050, but don't like the additional lag the HiDPI version introduces, you can set it to actually run at 1680x1050. Because of the small physical size of the screen the resulting image will still look pretty decent, but will be significantly less taxing on the system.
Cheers.