The equivalence principle tells us that in some local neighborhood, every free-falling observer in a general relativistic spacetime will measure the speed of light to be $c$; this literally means at a given tangent space, a light beam's 4-velocity is a null vector. Following the geodesic postulate, it will go on its merry way following a null geodesic.
But as we know, light beams once sufficiently far away in a curved spacetime will appear to travel at coordinate speeds not necessarily $c$. The canonical example is the speeds $\frac{dr}{dt}$ of radial light beams approaching zero as they near the event horizon in the Schwarzschild spacetime from the perspective of an observer at asymptotic infinity (and so when using the standard Schwarzschild coordinates).
But what practical experiment could give a value for $\frac{dr}{dt}$? One can't bounce radar signals off the beam, for example. And even then, how would $r$ values be found for the beam of light at a given $t$?
Ingredients: We have physical meaning for the $t$ coordinate as proper time for the asymptotic observer (call their worldline $\mathcal{I}$). And the $r$ coordinate gives the physical circumference for circle $t = 0, \theta = 0, r = R, \phi \in [0,2\pi)$.