I have a HD 6550M and it wouldn't wake up correctly with fglrx for several months.

There were different results when waking up:
1. nothing worked. Even the hard disk wouldn't spin up. And yes, it was fglrx because when booting with "nopat" fglrx used some own pat implementation and 1. didn't ever happen with nopat.
2. Just the graphical output being completely frozen and fglrx would write "ASIC hang happened" to the log.
3. X hanging with 100% CPU in R+ state with no error message.

Since I switched to the open source driver the notebook wakes up correctly about every time.

The xrandr compatibility was also kind of broken back when I used it: You need to have a X config with a Virtual Screen size big enough if you want to attach a second screen in before, otherwise you need to restart X (!!!) in order to attach a screen with its native resolution.
I also thought xrandr output like this was impossible until fglrx managed to report two preferred resolutions (even in wrong order):
CRT1 connected 1400x1050+1600+12 (normal left inverted right x axis y axis) 433mm x 271mm
   1400x1050      60.0*+
   1680x1050      60.0 +
   1280x1024      75.0     60.0  
   1440x900       60.0  
   1280x960       75.0     60.0  
   1280x800       75.0     60.0  
   1152x864       60.0     75.0  
   1280x768       75.0     60.0  
   1280x720       75.0     60.0  
   1024x768       75.0     70.1     60.0  
   1024x600       75.0     70.1     60.0  
   800x600        72.2     75.0     60.3  
   800x480        72.2     75.0     60.3  
   640x480        75.0     72.8     67.0     59.9  
   640x350        60.0     70.1
That said the topic was more about OpenGL compliance/performance.

There are some more topics one could discuss when it comes to sucking drivers. I am going to buy a notebook with a HD 7970M with AMD's optimus-like "Enduro". I'm not sure if this is a good idea but with GPU offloading already working it at least has a pretty good chance working in the not so far future if the proprietary AMD drivers fail. It will also be very interesting to see if the windows driver in a windows vm succeeds in using the dedicated card with vt-d and pci passthrough.