Any thoughts Pro/Con of operating a 400V 3-Phase 1200A Manual Transfer switch and how to safely implement this?

Thanks!

1. Has an arc flash hazard analysis been performed for this switch?
2. If so, was it an actual study or are you using the tables in NFPA 70E?
3. If using the tables, how do you know you are meeting the parameters of the table?

The above will allow you to select the proper PPE for this task. Since it is a manual transfer switch, it should be operated without any load on it to be safe.

Why??
There are many load break rated manual transfer devices.
I don't know why you would want to slam 1200A of load onto a source, but hey that is what happens with a regular ATS during retransfer.

Whenever possible it is preferred to transfer without loads, but certainly not a requirement. All switches are rated for a number of mechanical cycles, and electrical cycles under certain conditions such as under rated load or under maximum interrupting load. The number of cycles gradually decreases in this order. So for instance a switch might be rated to 10,000 mechanical cycles, 1,000 cycles under normal rated load, and only 1 cycle under maximum interrupting capacity. Also, some switches are not designed to break loads (non-LBS).

Note that these parameters are much easier to find with utility grade gear than others but the data is there if you look hard enough for it. For the record I've put in a lot of manual switches. Depending on conditions it is often easier and more reliable to go find an available generator or welder that works than to maintain a dedicated, fixed backup generator. I use an ATS though when it's a dedicated generator.

Yes, ATS will transfer under load but it is faster than a person manually operating a transfer switch. I think is safer to transfer with as much load off.
Question to Kimo: Does the transfer operation happen prior to generator starting or are the loads operating, generator operating and then transfer switch operating? You substitute second source for generator here.

Most load break rated Manual Transfer Switches operate as quickly as an ATS and typically much faster than M-T-M throw over schems using separate devices.

Based on the most recent information, I see nothing in the OP's system that could not be powered down so that the transfer is done no-load.

The compressors don't have a start button? Just an ON button?
Maybe a small relay timer in their control circuit when they see power would be enough to not all start at the same time when power comes back, with a different setting for each (staged). Just an idea.

Does your manual TS have in-phase monitoring? If so, will it be enabled? Because you might be transferring power with both sources operating and have motor loads, consider writing a method of procedure and or doing the transfer with the power off. If that's not always possible, see if you have in-phase monitoring in the switch. The idea behind in-phase monitoring is to prevent damage the equipment connected to the sources, it's also to prevent tripping breakers or fuses. ASCO says the inrush with an out of phase transfer can be very high.
The following comes from ASCO's website
"An In-phase monitor is a motor load transfer feature that monitors the voltage, frequency and phase angle of the two sources. It ensures the two sources are within predetermined phase angle before the ATS controller initiates transfer of the load to the alternate source. The transfer is completed in a safe window to limit the motor inrush current to be equal to or less than normal motor starting currents"

Just looking at your ratings, honestly, good luck getting arc flash down to reasonable levels. Strongly suggest that you consider putting in more transformers and distributing the load out more. For 480 V three phase, it is difficult to do much of anything with arc flash once you exceed around 1500 kVA. Once you go to a higher voltage level though (2300 or 4160), I'm not sure where the cutoff is since I've successfully kept it down to reasonable values up to 10,000 kVA which is the largest transformers that I've worked with in recent memory, even with very low impedances.