Leaving all the various regs e.g. BS7375 aside for the moment, this doesn't follow logically IMO. Over a large PME system, the many local earth electrodes back each other up and provide a distributed low-impedance connection from the CNE to real earth. But if failure of the incoming CNE to the site itself is a specific elevated risk, none of the other electrodes are available as a backup; only the local rod will be connected to the MET and all the site's neutral current will be diverted into it which could easily result in a high touch voltage. This would not occur with a purely TT system, where severing an incoming neutral conductor has no impact on the touch voltage and merely stops the loads working.
What the regs make more of a point of, is the near impossibility of bonding all extraneous metal on a site to the CNE. Thus the equipotential zone is reliant on conductivity via true earth and therefore the MET should be as close as possible to true earth, achieved by TT rather than TN-C-S even with a parallel local rod, because the Ra will be much higher than the Zs and the external system potential determined by the prevailing system CNE voltage.
AFAIK neither BS7671 nor 7535 explicitly prohibit TN-C-S but draw attention to the fact that it probably isn't feasible. E.g. 704.411.3.1 invokes the requirement for 411.3.1.2 which is the general use of EB. Then G12 from the ENA states that as this isn't usually practical, the TN-C-S earth shouldn't be provided on a site temp. Etc.
In summary, IMO the advantages of being completely disconnected from the TN-C-S earth and network conditions outweigh the disadvantages of relying on RCDs for fault protection due to the high Ra.