The explanations as given in post#8 can be easily verified by inspecting the corresponding real closed-loop gain function (for finite and frequency-depending open-loop gain properties):
Example: Non-inverting amplifier (negative feedback via R2-R1).
It is easy to compare both cases if we write the closed-loop gain function as a product (gain ideal)*(error function) which turns out to be V2/V1=(1+R2/R1)/D(s)
1.) Voltage opamp: D(s)=[1+(1+R2/R1)/Aol(s)]
2.) Current-feedback amp CFA: D(s)=[1+R2/Ztr(s)]
As can be seen, in both cases the closed-loop gain approaches (1+R2/R1) for ideal devices (opamp open-loop gain (Aol) and CFA transfer resistance (Ztr) approaching infinite, resulting in D(s)=1).
However, there is one important difference:
1.) Voltage opamp: In D(s) we see that (1+R2/R1)/Aol(s)=1/[(Aol*R1/(R1+R2)]=1/LG(s) ;( LG=Loop gain)
2.) Cuurent-feedback In D(s) we have R2/Ztr(s)=1/(Ztr(s)/R2)=1/LG(s)
In both cases the error function contains the finite loop gain LG(s).
However, in case 1 (voltage opamp) the loop gain cannot be selected independent on the ratio R2/R1 (that means: not independent on the desired gain), whereas in case 2 (CFA) it is only R2 that controls the loop gain.
As a result, voltage opamps mus be frequency-compensated in order to operate for all closed-loop gains.
In contrast, CFA units can be made stable by proper selection of the feedback resistor R2 - independent on the desired gain (1+R2/R1). For this reason, CFA units do NOT require any frequency compensation and, hence, provide a larger bandwidth.