I will have to check what is the time constant we used, but the 5000RPM is a so fast speed that I don't think it will be the issue, because it would cause problems on low speed when the spindle is rotating slower than the time constant of the measurements and this is sure not the case with the high 5000 RPM.
I think it is more likely that the problem is the opposite of that, that the index signal is so short that the controller can't read it.
A quick calculation: 5000 RPM is about 83.33 rotations per second and that would be about 0.012 seconds per rotation.
Assuming that you have an encoder with e.g. 1000 counts per second and that the index pulse is 1 pulse long only then one pulse is only 0.012/1000 seconds long which is only 12 microseconds.
The controller needs at least about 20usec to safely read the signal, because there is a small low pass filtering implemented mechanically on the inputs to avoid noise spikes to make false triggers.