What this means is that choosing another method to increase the gain might be better because it may not limit the usable frequency range as much. The transistors are much faster than the op amps, so increasing the current mirror current is probably a better method to increase overall gain. In fact, that may be one of the advantages of using current mirrors in the first place but you'll have to look into this more. If the output stage gain was set to 1 with an increase in gain somewhere else then we might actually see amplification by a factor of say 10 without impairing the frequency response very much. You'll have to try this because there is more involved than just setting the gain.
Thinking out loud, if the output slew rate is 1v/us then with a 10k output resistor on the first stage we get 100ua/us which leads to 100ua/us on the output and with a 10k on the output we see also 100ua/us which leads to 1v/us so the response did not improve.
If we decrease the buffer output resistor to 1k, we still see 1v/us but now we see 1ma/us, and so 1ma/us on the output and so with that same 10k output resistor we see 10v/us, so we've increased the slew rate.
This is all theory so far, there will be limits as to how far we can take this if in fact we did not overlook anything already.