Think of it like this: The wavefront around a dipole is comprised of near and far fields, with the near field being spherical and the far field planar.
That near field has a reactive component (stored energy) that does not propagate, but falls off at 1/r^3 or faster.
So an incoming plane wave induces charge motion, which builds up the reactive near field over many cycles, generating that spherical wavefront.
So that seems to indicate reciprocity is only valid for a steady state, but it’s still valid. If your transmit antenna were fed a monocycle (i.e. not time for the near-field to build up), the receiving antenna wouldn’t have enough time either.
Evanescent fields exist sure but Far field is really just a useful mathematical construct. It is typified by a wave-front where you can approximate no phase difference wherever it lands on a planar antenna.
It's like how you can approximate the Earth as flat when making a platting because it is very large and you are very small. If you look at the Farfield approximation calculation for a large antenna or phased array, you'll see that the equation is a function of distance, wavelength, and aperture size.
Edit: I should point out that Evanescent waves do not carry power (no net energy flow) so the power transfer is always reciprocal between 2 antennas.
That near field has a reactive component (stored energy) that does not propagate, but falls off at 1/r^3 or faster.
So an incoming plane wave induces charge motion, which builds up the reactive near field over many cycles, generating that spherical wavefront.
So that seems to indicate reciprocity is only valid for a steady state, but it’s still valid. If your transmit antenna were fed a monocycle (i.e. not time for the near-field to build up), the receiving antenna wouldn’t have enough time either.