About at this point, even the best superconductors, at LHe temperature, can't sustain operation. These two cases, and many at lower intensities (including thousands of MRI machines in the 1-5T range), are all superconducting.
Superconductors are a big win, because the field isn't so high that it's impossible, and because most of these require tons of total energy (it's intense /over a large volume of space/, too).
It's impractical to try and charge one of these things every time you need to use it, let alone run steady-state, while burning a tremendous amount of power in a resistive material. You'd have problems with mechanical expansion, thermal gradients, circulating fluids (I mean, you need LHe and LN2 otherwise, but not at the pressure and flow rate that you need for the cooling fluid to run a copper magnet this big) and so on.
...These ones, however, may not be superconducting! Or at least not wholly, AFAIK.
My familiarity with critical field is probably lacking as well; I just remember it's somewhere in that range.
You'd have to ask the designers of that particular NMR. It might be feasible: after all, magnets can be stacked in a particular way so that their vectors add up (Halbach array and such). They might be using a similar trick to optimize the use of superconductors.
The strongest are either fully resistive, or hybrid. The ref shows how this one is built:
And note that boosting a 20.0T superconducting magnet, up to 21T as a hybrid magnet, takes exactly as much additional effort as making a 1T air core magnet does (not easy!). So, to achieve another 20-some T extra from that resistive magnet is no small feat!
Tim