Weird clock behaviour on dataread (AT91SAM9260)

I have a 6 layered sam9260 board and I'm seeing something strange on
the 18Mhz clock.

This is a scan from the SAM9260-EK (reference) board, when there's any
data written to the dataflash, a probe on the crystal shows this
pattern. (notice the slightly raised ground, which then stabilizes and
continues.

[URL=http://img121.imagevenue.com/img.php?image=44891_ek-
normal_122_731lo.JPG][IMG]http://img121.imagevenue.com/loc731/
th_44891_ek-normal_122_731lo.JPG[/IMG][/URL]

here are a couple of scans from my custom board.
Similarly when writing to the dataflash, the clock goes through this
weird cycle. This ultimately results in an inability to communicate
with the dataflash.

[URL=http://img180.imagevenue.com/img.php?image=44898_DUT-
sngl_122_493lo.JPG][IMG]http://img180.imagevenue.com/loc493/
th_44898_DUT-sngl_122_493lo.JPG[/IMG][/URL]

this is a faster scan of the same clock, but with a capture of one of
the SPIO data lines on ch2.

[URL=http://img102.imagevenue.com/img.php?
image=44899_DUT-2chnl_122_872lo.JPG][IMG]http://img102.imagevenue.com/
loc872/th_44899_DUT-2chnl_122_872lo.JPG[/IMG][/URL]


Bypass capacitance was initially suspect since I don't have as many
10uF tantalums on my board as the EK does, but even a 220uF tantalum
across 3.3V didn't seem to solve this. There are enough (as many as
the EK has) 0.1uF bypass caps for both the micro and dataflash
though.
Strangely a probe directly on the ground plane doesn't seem to show
any fluctuations during the cycle.
I am considering possible heat damage to the micro, and am having a
board made by the fab house instead of my toaster oven.
However I wanted to see if someone could explain this weird spike in
the clock (even in the EK board) and how you think they deal with it,
that my board can't fix despite having the same schematics.