The last time I visited this issue, which was last year with a 680i board and a B3 quad-core Q6600, I could see how the Speedstep feature could be an instability problem (and correct me if I'm wrong, but this is what I observed.)
The CPU is set to run at a certain multiplier and external-frequency (or what is known as CPU_FSB). The Speedstep feature works by dropping the multiplier, and the multiplier on any given processor below the flagship model will only be an integer value in a fixed range, the highest value being the default multiplier.
If you over-clock by lowering the multiplier and raising the FSB, the Speedstep feature may attempt to drop the multiplier below the minimum value. At least, this is the way I remember it, and I believe (and believing isn't knowing, and suspecting is between either one) -- that this explains it. So I "suspect" this to be the explanation.
IF SUCH IS THE CASE, then over-clocking with the stock multiplier may "play nice" with the Speedstep feature, anyway.
That's my take on it.