Representing angular velocity using quaternions

If we let:

a(t) = a0 +w * t

where:

• a(t) = angle in radians that a solid body has turned through at a given time about a constant axis
• a0 = angle at time t=0.
• w = angular velocity in radians per second about a given axis.
• t = time

The quaternion representation of a rotation angle (orientation) is:

q = cos(angle/2) + i sin(angle/2)*axis_x + j sin(angle/2)*axis_y + k sin(angle/2)*axis_z

where:

• axis_x = component of axis in x dimension
• axis_y = component of axis in y dimension
• axis_z = component of axis in z dimension

so if the quaternion is a function of time we substitute angle for a0 +w * t to give:

q = cos(( a0 +w * t)/2) + i sin(( a0 +w * t)/2)*axis_x + j sin(( a0 +w * t)/2)*axis_y + k sin(( a0 +w * t)/2)*axis_z

So this gives the orientation as a function of time, so to get angular velocity we need to differentiate this with respect to time, as discussed here, so we differentiate q to get qw which is a quaternion representing angular velocity.

So if we differentiate each term we get,

real terms ---> d(cos(( a0 +w * t)/2))/dt = d(cos(( a0 +w * t)/2))/d( a0 +w * t)/2) * d( a0 +w * t)/2)/dt = -sin( a0 +w * t)/2)*w

imaginary terms ---> d(sin(( a0 +w * t)/2))/dt = d(sin((a0 +w * t)/2))/d(( a0 +w * t)/2) * d(( a0 +w * t)/2)/dt = cos((a0 +w * t)/2)*w

So if we put the quaternion back together we get:

qw = dq/dt = w * (-sin(( a0 +w * t)/2) + i cos(( a0 +w * t)/2)*axis_x + j cos(( a0 +w * t)/2)*axis_y + k cos(( a0 +w * t)/2)*axis_z)

So what is the physical interpretation of this? Unless I have made an error (which is very likely) an instantaneous rotation velocity can be represented by a quaternion but its elements would be continuously varying with time in a way that is difficult to interpret?