David, thanks for the clarification.
Not to bore you with this, but I took the time and loaded that circuit into LTSpice. I did the best I could with the models and it looks like, according to LTspice at least, that this circuit should have significantly higher bandwidth than what we are seeing. I see the correct output into 50 homes of .2 V and .4V into 1Meg ohm and I see some rise time and fall time increase at 5Mhz (in my model) but nothing like the distortion we see in the live circuit.
When I probe the live CA3046 amp at for instance pin 10 and 7, I start to see high frequency drop off adding rise time of about 40ns to the 10ns in the square wave from the time base. I then see even more rise time (and fall) added as it goes through the next transistor so that once it hits pins 8 and 6, it is pretty much a 5Mhz triangle at that point.
I thought I had some CA3046 chips around here and if I find one, I might breadboard the circuit to see what maybe a new version of that chip might produce.
Again, thanks for the pointers and clarification.
Jerry