Everything about transmission line theory is a useful approximation, not a detailed description of physical reality.
When transmission line effects are important, it's only an approximation to assign a scalar potential to any point in the circuit. Whether on the transmission line, on the ground plane, or at the terminations. Nonetheless we approximately define the scalar potential, and we find this is useful for predicting the circuit behavior at a high level.
In the microstrip transmission line, the actual behavior comes from the electromagnetic wave travelling between, and (approximately) confined by the boundary structure formed by the ground plane and the trace. We can define an (approximate) potential at each axial location along the line by the integral between the ground plane and the line (along the shortest possible path) at that axial location.
It isn't important whether the potential varies between different points on the ground plane or whether the potential varies between different points along the trace. It's only important that the (approximately defined) potential difference between the trace and the ground plane varies along the length of the line. Since this is all that matters, it's useful to assume that the ground plane remains an equipotential and that all the potential changes happen on the trace.
From experience we know that this approximation is close enough to reality to allow us to predict the circuit behavior as observed at the terminations of the transmission line reasonably accurately.
In summary: You are correct that in a high-speed circuit there will be potential variation between points on the ground plane. Nonetheless we make a useful approximation and assume the ground plane is at a uniform potential and we find this approximation is useful for predicting circuit behavior.