There's no doubt whatever that the speed of light is isotropic around anything fixed with the ECI frame. Ring laser gyros demonstrate this fact conclusively.

The nodes of wave reinforcement in a large enough ring laser gyro oriented perpendicular to the earth's axis of rotation and centered on that axis, will remain fixed with the star background. If it were possible to expand the ring diameter so that it encircled the earth around the equator, the nodes would still remain similarly fixed. There's no doubt about that. Nor is there any doubt that the nodes would pass by anything fixed with the rotating earth's surface at the rate of 465 m/sec, and the only way that can happen is if the speed of light around that fixed point is 465 m/sec slower than it is in the ECI frame. The speed of light is (local c) + 465 west, and (local c) - 465 m/sec east.

But none of that can be applied to SR of course because the rotating point is not inertial. The same can be said for GPS satellites, so why is it applied there?

The straight line path taken by an inertial traveler moving at 465 m/sec in an easterly direction relative to the earth is at a tangent to the earth at the equator and coincides with the straight light path section between one set of mirrors in an ECI frame fixed, equator encircling gyro. From the viewpoint of the traveler, relative to the earth the nodes of wave reinforcement will still be seen to be moving to the west at the rate of 465 m/sec. The only change in the latter scenario is the sideways shift of 7.8 centimeters per kilometer required to change the curved path to a straight line trajectory. How could that possibly make any difference?

The laws of the universe abruptly change from the simply logic of c'=c+v and c'=c-v, to physical length contraction of the ECI frame. That is nothing more than unsubstantiated waffle.

Many years ago I set about trying to detect an expected 400 m/sec east and west light speed anisotropy. I was surprised by the null result when I turned this apparatus between east and west.

The expectation was that the vastly different number of waves along the air and glass paths would not change in exact proportions as the device was turned. But exactly the same number of waves were apparently being added to or subtracted from each path.

The latter part of the link at the bottom of the page explains the null result.

One means of generating an asymmetry in that system would be to somehow set the glass path in motion along the line of the beam, i.e. a spinning glass disc. But generating a stable interference pattern would be impossible because one could hardly even breath anywhere near the apparatus without sending the pattern into a state of chaos.

Some years later the Fizeau ether drag experiment began to ring some bells. That device was able to maintain a stable interference pattern while liquid was being pumped around the light paths at fairly high speed. The reason for this wasn't too hard to understand. The light beam is split by a semi silvered mirror and the two halves are sent in opposite directions around almost exactly the same path. The forces distort each path equally so it's almost impossible to change the length of one path and not the other.

Thank you Mr. Fizeau.

This is the updated setup.

Variable water flow is toward the laser end.

The system would be expected to phase lock using such a light source because there is feedback directly into the exit window of the laser. It does exactly that, but it doesn't involve the beams traveling through the water. The most intense reflection is generated at the entrance windows of the water tube and that is the logical limit of the phase locking loop. Two interference patterns arrive at the screen, but the much weaker signal is not phase locked.

The main purpose of the experiment was to detect a fringe shift, not to precisely describe how much it would be. Since nothing is changing while the device is turned between east and west, the only possible errors will stem from friction in the water connection, and leveling of the apparatus. But since both paths are always equally affected, the result would still be sound.

The gravitational distortion of the whole apparatus when it's turned from horizontal to point vertical causes a fringe shift of only three wavelengths. Even a one degree leveling error wouldn't change the result enough to cause a problem.

The experiment Light speed anisotropy proven was for obvious reasons done on a shoestring budget, but the result is sound.