The main purpose for blind testing is to identify the audibility of effects that are already known to be measurable. For instance it can be used to determine thresholds of audibility for effects such as distortion and lossy data compression. A lot of what we know about hearing is due to blind testing in a research setting. For instance, I suspect that the Fletcher-Munson curves were determined in this way. A lot of blind testing can now be done with just a computer, as the test signals can be generated digitally.
The current proposal on the table is to use blind testing to settle a controversy over an effect that is not known to be measurable. To make blind testing work, the cable has to be the only difference under test. Adding a speaker, microphone, and acoustical environment simply raises the risk that other artifacts will creep into the recordings, such as room noises or the length of dead air at the start of each sample. If it isn't all done just right, nothing will be learned. At best, any result is only likely to be regarded as a preliminary step towards additional corroborative testing.
This is why I'm not hanging my hat on blind testing to settle this issue.