As far as an amp is concerned, a higher impedance load is an easier load. This is also why pretty much every preamp out there can drive a 10kOhm input impedance into an amp easily, but you have to have some pretty serious professional (or high end) amplifier to push any kind of substantial voltage swing into a 2 ohm loudspeaker. Since the amp is trying to act as a voltage source, it isn't going to clip if the impedance of the headphones is higher in a particular area - it just doesn't have to source as much current at those frequencies for the same voltage swing. The O2 can easily drive the HD800 well past any volume you'd ever want to listen at, and it will do so with zero audible distortion and perfect frequency response. As I mentioned earlier, the O2 can push around 7V RMS into a high-impedance load like the HD800 (and if you look at the designer's graph, you might notice that its voltage output vs THD plots look almost identical when driving an 80 ohm, a 150 ohm, and a 600 ohm load, and its THD drops and peak output voltage increases for higher impedance loads). I know people in high-end audio like to believe that a $150 amp couldn't possibly drive a $1500 pair of headphones correctly, but the measurements disagree in this case. I will admit though that my prior calculation is wrong - I forgot that Sennheiser specifies sensitivity in dB/1VRMS, not dB/mW like everyone else, so a 7.3V signal would actually only push them to 119dB, not 125 as I stated above. Of course, 119dB is still way more than enough.
As for the HD800 vs HD650? Assuming the manufacturer's specified sensitivities are correct, there really shouldn't be much difference between the difficulty of driving the two of them. Sure, the HD650 is a bit more sensitive, so it should be a bit easier to drive, but the difference should be on the order of a dB or two. Unless you were already really, really borderline with the 650, the 800 shouldn't be a problem for any amp that can drive the 650. How are you determining that it is clipping, out of curiosity?
Finally, as far as the amp is concerned, the most difficult load to drive well is actually a low impedance, low sensitivity headphone, such as the HE-6. To drive those properly, you need a large voltage swing and a LOT of current, and that is where you could well see a problem with the O2 distorting. It does just fine on the LCD-XC in my opinion, which is the hardest headphone to drive that I have ever tried my O2 with. The HE-6 is even harder to drive though, so I could envision it having some problems there. I don't have access to a pair to try them out, so this is purely theoretical. Actually, in many ways a moderate sensitivity, high-impedance headphone is the easiest to drive (which is probably part of why many professional monitoring headphones fall into this category). A super sensitive, low-impedance design like an IEM is really sensitive to noise in the amplifier circuit (which can be hard to eliminate entirely), a low impedance, low sensitivity design requires a lot of current (which can be hard to source without distortion), but a moderate sensitivity, high impedance design is not going to have a noise problem (unless you have a truly appalling amount of noise in your electronics), is not going to require much current to drive (which means things like output impedance and trace resistance on the board become unimportant), and the only real downside is that it takes a bigger voltage swing (and +/- 20V or so is easy to do when you have access to an AC power supply - it only really is a problem if you have to run off batteries in a portable player or something like that). So, if you know you'll pretty much always be using a desktop amp with high voltage power sources available, the high impedance, moderate sensitivity design is kind of a no brainer. I don't know where people got this idea that high impedance = hard to drive, but it's really not even close to correct (sadly, otherwise I'd be running my 8 ohm speakers off a tiny little headphone amp
)