Cosmology: Is there any experimental evidence for the redshift scale-factor relation? Modern cosmology relies heavily on the redshift scale-factor relation $$a=\frac{1}{1+z}$$
But what is the experimental evidence for it?
It's derived in textbooks from General Relativity, but for all other predictions of GR there have been stringent tests.  The bending of light by stars, precession of the perihelion of Mercury, the equivalence principle, time dilation, etc... have all been subject to high precision tests.  Even if GR is correct, it's possible that the derivations misinterpret something...
Have any tests been done, or are any planned, that that look for direct experimental evidence of the relation?

The question is different to this one What experiment would disprove Friedmann model of cosmology? as it asks specifically about attempts to experimentally verify the redshift scale-factor relation.
 A: First of all, the scaling law is as follows. For a photon detected emitted at $t_0$ and detected at $t_1$ we have the redshift $z$ of the photon given by the scale factors $a$ at times $t_0, t_1$ as
$$\frac{a(t_0)}{a(t_1)} = \frac{1}{1 + z}$$
That is, the redshift is derived from the ratio of scale factors. The issue with confirming this relation is that $a(t)$ is a metric component, and metric components are typically not directly observable. One has to be very careful - what exactly is meant by "testing" the $a-z$ relation? What are the observations that we are comparing?
Let's take a look, the scale factor $a(t)$ is special, since it chosen to correspond to the rescaling of distances in the cosmological comoving frame. The $a-z$ relation could then be taken as a definitory statement, since it really corresponds to the stretching of distances between the wavecrests of the photon. In that case, one would just need to verify the isotropy and other assumptions of the FLRW cosmology to test whether such a definition of the scale factor is self-consistent. This is actually how cosmology is built in practice, astronomers essentially do not refer to $a$, they refer to $z$ and define the cosmological eras using the redshift as a "coordinate".
Another operational definition of $a(t)$ (up to a rescaling) could be also from the average distance between ordinary matter such as galaxies and clusters of galaxies. If one is able to track all the matter in one era and measure its distances, compare this with another era, one essentially finds the ratio of $a(t)$s between these two eras. Does this ratio agree with the redshift observed between them?
Unfortunately, this simple idea is extremely difficult to realise. At short distances you can cover things well but you do not have the statistics (you need Gigaparsec volume averages for FLRW cosmology to start making sense), at larger distances you start having selection effects (for instance, you notice less bright things less). And when you pass between eras with very different redshifts, you have different selection effects for every era: you use different instruments with different capabilities because the wavelengths become different (and you need different sensitivities), and even the objects themselves change, since they may be glowing more or less dependent on the era you are in. In short, there is no good experimental test of a definition of $a(t)$ based on average distances, you have to take the redshift relation as more of a definition.
A: The point here is that redshift and scale factor are the same thing. If you assume that light speed is constant everywhere, which is a reasonable well tested assumption, then the correlation between scale factor and redshift follows.
Imagine two galaxies A and B. Imagine to send a a laser beam with a certain wavelength. Now imagine to divide the distance between A and B by  and call this number n (number of wavelength contained between A and B). If n is fixed, stretching the space between A and B will give you a different value of . As you may see if space is stretching the wavelength varies.
A: Edwin Hubble discovered the law which bears his name in the 1930s.
By determining the apparent brightness of a star and its distance away, its intrinsic brightness or luminosity can be calculated. At the time of Hubble, the distances to many nearby stars had been calculated, for example through measurement of geometrical parallax from different observation points.
Stars of a type called Cepheid variables pulse steadily in their luminosity. Using the above method, they had been found to have a direct correlation between their pulse rate and luminosity, so by measuring the pulse rate their luminosity could be determined.
This allows the distance of a far more distant Cepheid variable to be determined, simply by measuring its pulse rate and apparent brightness. Such stars can thus be used as standard candles to determine how far away they are.
Hubble correlated these calculated distances with their measured spectral redshifts, to discover their direct relation.
His final step was to interpret redshift as indicating a receding velocity (a phenomenon already known), leading to his discovery that the Universe is expanding. But that goes beyond the redshift-scale factor relation asked about here.
Subsequent measurements, using these and other standard candles, have confirmed and refined his results.
A: In the context of FRW cosmology, the equation of you wrote is just the definition of the variable $z$. It is straightforward to show that in the FRW metric, as light travels, its wavelength is stretched by the scale factor $a$, which gives a redshift.
But I see that you are looking specifically for experimental evidence. That's tricky, because the scale factor $a(t)$, or $a(z)$ is a property of the FRW metric. The FRW is an meant to be an approximate model for our universe, where the universe is homogeneous and isotropic. Obviously, that is not the case on scales less than a Mpc, so the "scale factor" doesn't strictly exist.
On the other hand, you can define an approximate scale factor on large scales. We usually measure it in terms of redshift, i.e. we assume FRW because this is the model we are testing.
But I suppose that if there were an object whose size were known, we could compare its apparent angular diameter to its intrinsic angular diameter, whose ratio should be $a(z)$. We could then compare the apparent wavelength of its light to the intrinsic wavelength of its light, whose ratio should be $1+z$. I guess that would be a test of $a=1+z$. But this sounds tricky. I'm not aware of such a test being done, but maybe somebody knows one?
