Too long, didn't read answer: Abuse of notation, reference frames, and the transport theorem.
The left hand side of that equation would be better written as ##{\frac{d\vec L}{dt}}^{(I)}##, where the superscript (I) means "the time derivative of angular momentum from the perspective of an inertial frame."
Without proof, the transport theorem relates the time derivative of some vector quantity ##\vec q## as observed from the perspective of an inertial observer versus that of a rotating observer as
[tex]{\frac{d\vec q}{dt}}^{(I)} = {\frac{d\vec q}{dt}}^{(R)} + \vec\omega \times \vec q[/tex]
Use ##\vec q = \vec r## (i.e., position vector) and differentiate twice and you'll get the standard relation involving centrifugal and coriolis accelerations between accelerations as observed in a inertial frame versus that observed in a rotating frame.
What happens if you use ##\vec q = \vec L \equiv I\vec \omega\ ##? The left hand side, ##{\frac{d\vec L}{dt}}^{(I)}##, becomes the external torque. The right hand side becomes ##{\frac{d\vec L}{dt}}^{(R)} + \vec\omega\times \vec L##.
In any frame, ##\vec L = I\vec \omega##, but you have to beware that the inertia tensor I and the angular velocity ω are frame-dependent quantities. The inertia tensor of a rigid body is constant in a frame rotating with that body. Thus ##{\frac{d\vec L}{dt}}^{(R)} = {\frac{d}{dt}^{(R)} (I\vec \omega)}= I\frac{d\vec \omega}{dt}##. Note that I've dropped the superscript R from the final derivative because angular acceleration (time derivative of angular velocity) is the same vector in the inertial and rotating frame.
Putting this all together yields
[tex]
\vec \tau \equiv \frac{d}{dt}^{(I)}(I\vec \omega) =<br />
I \frac {d\vec \omega}{dt} + \vec \omega\times(I\omega)[/tex]