The GR Research guy (along with some others in history) already showed the frequency response doesn't change much if at all. He also claimed the difference was very audible despite the measurements matching.
He made a set-up where one speaker has two crossovers inside. You can switch between them. If anyone has the gumption to try this, then that would be an excellent thing.
The main thing is that there are people saying measurements do not tell the whole story, so you therefore can not use them in this debate. It must ultimately be a listening test to determine if there are things happening that can not be measured (or measured as of yet)
Danny Richie from GR-Research will disagree with this review.
I think I once saw a video of "newrecordday" or something similar on Youtube.
I missed a few critical questions there, e.g. were the replaced parts measured and did the new parts have exactly the same values? As a rule of thumb, at least one digit behind the indicated component value - e.g. old 4,735 ohm, new 4.732.
As already written in the review the tolerances of sand cast resistors are usually 5%, which is clearly audible when installed in series to the tweeter (You will find the reasoning below).
The same applies to the audibility of component tolerances for capacitors installed in the tweeter circuit.
"The main thing is that there are people saying measurements do not tell the whole story..."
This is also the reasoning of the cable sound and capacitor sound fraction in the audio world. There it is mostly about previously undiscovered electrical phenomena.
That's the reason for the experimental setup with the tweeter and the measurement with the measuring microphone (I used the same setup for my condenser measurements, if there is time I'll post them in the forum).
To my knowledge there is nobody who would seriously claim that there are undiscovered possibilities of sound transmission (from tweeter to microfon).
Or that an exchange of resistors changes the radiation of the tweeter.
So the problem is reduced to the question "
When do the measured effects become audible".
Here is a quote from Toole's book "Sound Reproduction
The simplest deviation from flat is probably a spectral tilt. There is some evidence that we can detect slopes of about 0.1 dB/octave...
So let's have a look at what can happen when a resistor with +-5% tolerance is replaced.
A resistor with 3.3 ohm can have the real values 3.14-3.47 ohm (0.33 ohm tolerance), a 6.8 ohm resistor 6.46-7.14 ohm (0.68 ohm tolerance).
We don't want to exhaust the limits completely, because this is not very likely, so we simulate with 0.27 and 0.56 ohm deviation. Let's simulate what happens when an ideal loudspeaker with LR4@2kHz crossover gets a slightly different value by replacing a resistor:
In both cases the change in slope is 0.1 dB or more per octave (e.g. from 2kHz to 4kHz). So the changes should be audible.
If you consider that there are also sand cast resistors with 10% tolerance, it should be clear that if you replace them with a resistor with 2% tolerance, it is very likely that there are audible changes - but these are only due to the different component values.
I shudder to think how many DIY projects may have under-rated resistors - possibly even commercial ones.
A friend of mine likes to listen his DIY hifi speakers very loud (he uses 2x1000W@4ohm 0,06 % THD+N).
This is what a 5.6 ohm 20W sand cast resistor serial to the midrange driver looks like afterwards, the applied hot glue is completely charred
...but the resistor survived it, nevertheless we changed to 2x20W sand cast resistors without using hot glue to fix them