ESP is an acronym for Electrical Submersible Pump (probably because it sounds better than "down hole pump" when speaking in mixed company). He's right, the XFMR -VFD -XFMR method was very common for that industry, mainly for two reasons:
1) The applications were always pumps, which means they did not need Sensorless Vector Control. One downside of doing it this way is that the VFD cannot use a motor model that remains accurate because of the added impedance of the transformer, which changes dynamically with frequency. The motor does too, it's just that the doubly dynamic interactions are too complex for SVC drives to track. But for the most part, you don't need SVC on a centrifugal load like a pump, so having to put the VFD in V/Hz mode is no big deal.
2) The ESPs were sold as an engineered packaged solution, so the step-up transformer design issues were dealt with by the vendors, who were looking for a cheaper solution than MV drives. And yes, there are several concerns for the transformers used. The ESP vendors initially went with the transformer mfrs who suggested specially built core designs to deal with the issues of operating with variable frequencies and high harmonics. But as more vendors got into it and pricing pressures increased, some of the low end ESP vendors started using off-the-shelf transformers, knowing they would likely outlive the warranty anyway, but leaving the future mess to the end users. But for a long time, MV drives were only available from a few vendors, so the ESP packagers could not negotiate pricing as easily as they liked. That's no longer the case, so the ESP vendors are tending to move away from that practice.
The argument the ESP pump packagers used to allow it to appear less expensive however was that the load side transformer was not a transformer, it was a part of the motor circuit. Therefore, it was being protected BY the same motor protection inherent in the VFD and needed nothing else. Not all inspectors agree with that however, and if you have to add primary and/or secondary (MV) protection and disconnects to that load side transformer, it immediately becomes economically unviable. If you use fuses for the protection, you also introduce the possibility of only one of the fuses opening on the load side of the VFD, not something the VFD mfrs. are thrilled about. In fact, they are not thrilled about opening a CB on the load side either, but it's at least acceptable in an emergency.
Probably the most overlooked issue is however, that there are added losses in that secondary transformer which are now permanent in the motor circuit, and the loss percentages increase as speed decreases (although net kW losses decrease because load decreases). It just means that the energy that you can save at reduced flow is less than if you went directly from a MV drive to the motor. Since this is typically only done on large motors, the $$$ add up quickly. The counter argument is that if you have a load reactor on the output side, it's the same as having a transformer. But that's not really true, because the reactor is simply magnetic, so the losses through it are very low. A transformer has added winding ratios, so the I^2R losses are added to the equation as well. You have all that extra winding wire now on the low voltage side added to the theoretical circuit length.
The issue of MV drives only being able to be serviced by MV technicians has also gone away for the most part. It was a problem initially, but several of the major players redesigned the drives years ago in such a way as to allow anyone to service them now. Some states require MV termination certification, but that will be the case no matter what, because if you have a MV motor, you will have to terminate MV cables somewhere.