I would suggest that you approach this very systematically. You community college should have rules in place to ensure facility safety, these standards should also have substandards for equipment.
If NOT the case, please conduct a proper risk assessment. From a machine standpoint, you should be able to put a safety circuit as a supervision which doesn't allow your students to make a mistake:
Potentially the EStop circuit can be the 1st step, then the next step can be motion. Perhaps you can also have something like a deadman switch to turn off motion, and supervisory max speeds to simulate a bench-environment.
Maybe you should consider pre-made trainers, or at least copy the features of trainers like guarding the rotation area, making the HV lines quick disconnect and insulated, just hedge the risk.
PPE is an important control, but elimination is the 1st issue you need to solve.
https://en.wikipedia.org/wiki/Hierarchy_of_hazard_controls
Also, you should have a circuit pre-evaluation in your class, and your students should not be able to energize anything before your checklist has been checked.
You can also take this opportunity to teach people about hazardous energy control including designing LOTO capable machines - even for testing.
Source: Had a "Collaborative" Universal Robot try to kill me when I changed the speed from within the AB PLC's UDT from What I thought was really slow 0.2 -> 5%.
In fact, 0.2 in the UDT represented 20%, and 5 somehow worked as a memory indirection and I remember reading the teach pendent went to 1400% of the max allowable UR speeds.
Moving faster than you can even move it manually.
Way beyond the Teach Pendant's allowed memory. The UR Killerbot was in fact not respecting it's own supervisory limits. I was saved by a workbench support beam, otherwise it woud have been a bad day for me testing this robot in my controls lab.