John, The input stages of the Allen Wright RTP (shown in the schemtic posted by Hiho) and of the Atma-sphere MP1 are both dual-differential cascodes , i.e., balanced topologies. Thus in both cases, a CCS is used optimally at the junction of the cathodes of the two "bottom" tubes in the cascodes and ground. (Take a look at the RTP schematic; I think AW used an LM317, which is definitely not my choice, but it's a CCS nevertheless.) Thus there is no place for more CCSs on the plate side. However, your point is relevant to a single-ended topology. But in a single-ended topology, the output Z would be the parallel sum of the impedances of the CCS and of the cascode stage below. The impedance of the CCS is ideally infinite (but is always finite in reality, albeit very high). However, would not the impedance of the cascode below be a very finite number, thus dominant in the calculation of output Z? And also subject to change as the tube(s) in the cascode age? Please correct me if I'm wrong; you probably know more about this stuff than I.
Now I re-read your post, you seem to be talking about a dual-differential cascode with CCSs between cathode and ground AND between plate and B+. AW discusses that possibility in his TPCB. Have you ever tried that?