I think it's the
full diameterfull Circumference=half circle circumference
Hi JP. Whoops, that equation looks like π = ½C, which is true, but only for unit circles! (Radian measure works with all circles.) It does seem like your intuition improperly associates rotations with pi in a one-to-one correspondence, but I'm still not sure why.
Regardless of the above, let's return to the radian unit-definition, referred to by Mark (i.e., in terms of arclength and radius) and see where that takes us.
Like π, the unit we call 'radian' is dimensionless and defined as a
ratio of two lengths. That ratio is [distance along the circumference] to [radius], or:
Angle measure in radians = arclength/radius
If the arclength subtended by an angle θ equals the radius, then the angle measures 1 radian.
θ = r/r = 1 rad

How many radii fit in a circle's circumference? A little more than six. In other words, for a complete rotation (arclength=C), we have:
θ = C/r = (2π∙r)/r = 2π rad ≈ 6.28
For half a rotation, the arclength is ½C.
θ = (½C)/r = (½)(2π∙r)/r = π rad ≈ 3.14
So, a smidgen more than 3 radii fit within half a rotation.
Everything is proportional, so we're not restricted to unit circles. For half a rotation when r=5, the arclength is roughly 15.70796.
θ ≈ 15.70796/5 ≈ 3.14159
θ = ½(2π∙5)/5 = 5π/5 = π
Therefore, the reason why π rad corresponds to half a rotation (versus complete rotation) is simply because that's the way the radian unit HAS BEEN DEFINED (caps for emphasis, not shouting).