I think in an ideal world, in a tube mike, one would send B+ and B- (power supply high voltage positive and negative) and H+ and H- (heater DC supply positive and negative) separately through the cable, to terminate into the power supplies analog 0V node. The shell of the mike body would then connect to the microphone cable shield, which is common with the PSU's metal chassis, which connects to the PSU 0V point in the PSU. In theory, a large radiated noise source can sum with the B- and be amplified inside the mike frontend if the chassis/shield and B- are intermingled.
However in practice, compromises are made: for example, with DC heaters, H- and B- are connected together without issue. In many designs, the mike shell is connected to B- as well, without issues as well.
In the C12 design from Alctron, I experimented with breaking the mike shell/headbasket from B-, and couldn't measure any audible improvement, even with the mike shell placed close to EMI sources like ballast's for neon tubes, CRT monitors, etc. I think the trade off between reduced susceptibility to EMI and minimizing ground impedance is probably hard to measure in that specific case.