Why do so many headphones spike at 10kHz?

As long as the spike's not too strong, I'm fine with it.  =p

I remember Tyll explaining the peaks/valleys in that region as being resonances between the driver and the ear, the concha, and inside the ear canal itself.  While the driver is could be putting out a flat frequency, the resonances make the measurements appear as if it isn't.  Apparently minor position shifts on the dummy head can cause quite a bit of variability in those frequencies as the position of the ear relative to the driver has shifted, thus altering the resonance.  

I think all this info is in one of the innerfidelity articles if you look at the back catalog.

I think the problem doesn't happen as often with open headphone. With the hd598, seems even the contrary problem.

My guess is that it gives a little bit of sizzle at a frequency that young and old can hear. If you had most people compare a flat response headphone to one with a bump at 10k I bet most would think the bumped one sounded better.  Consumer speakers often have the same thing.

No idea why, but that has been one of my biggest pet peeves since I got into headphones. Glad somebody else noticed, haha. I always have to EQ those spikes down if they are there.

Also, I've noticed that the spike is usually centered around 9kHz.

What looks like a spike is actually a more gradual increase if you make all of the increments on that chart evenly spaced. 

They're already spaced properly.  The graphs account for musical octaves.  As for why not 6khz, I thought it was a common practice among the audiophile headphones sennheiser offers to peak at 6khz-- right in the region of sibilance :S