Not really, the adage is a simplification. You can put metal in a microwave with understanding and consideration (or to recreate a known experiment from a reputable source, such as ones with a CD, light bulb or CFL) but it's easier to just teach people it's a no-no.

In fact, metal mostly reflects microwaves and only heats up substantially if there is a thin enough layer and/or nothing else to absorb the waves. Lone cutlery might get hot but like this, there is nothing the forks will ignite.

Thin metal layers heat up quickly. Aluminum foil and gold-plated china will spark and sublimate nastily, and this can of course ignite some dry food.

Metal containers shield the food from microwaves and disrupt standing waves, resulting in uneven heating so they should be avoided.

Closed metal containers (even meshes finer than the wavelength) prevent microwaves from reaching the food at all.

If there is no (reachable) food with water molecules, the cca 1 kW of microwave power has little to be absorbed into. The amplitude of standing waves inside will grow very high, resulting in more interference than usual. The waves will become strong enough to cause the 1 kW to be absorbed by the (not 100% reflective) chamber and magnetron and heat them up significantly. There is a resettable (bimetallic) thermal fuse (usually on both) but it may not be fast enough, in which case the magnetron's big ferrite ring will usually crack and/or lose its magnetism first. Some plastic parts (tray wheels) might melt or start smoking but even if the thermal fuse isn't enough to prevent them from catching fire, there is pretty good insulation to the outside and majority of flammable parts. I think microwaves are required by law to pass something like a "running empty for max time without catching fire" test even though it always says not to run empty in the manual.

Basically, think about it as a box made of mirrors with a 100W bulb hanging off the top. If there is a plate of food, it will get warm. If there is none, the chamber will get much brighter because of the near-infinitely reflecting light but the 100 W will have to dissipate somewhere, and even the infinite-looking mirror tunnel gets greener and darker. It takes dozens of reflections to dissipate most of the light energy, as opposed to 1 or 2 when there is a plate of dark food in the way.

I know about a fast food stand where the operators would just set the knobs to long time and high power, timing the processes with an external timer also used for frying etc. They would often close the door when done, resulting in the microwave running empty for minutes at a time. This wastes power and several of them got destroyed each year, and a penalty on running one empty was imposed as a result. Unfortunately, detecting the problem to allow for automatic shutdown that's faster and less prone to false positives than a thermal fuse is difficult, for example weight sensors can be cheated by a plate. Maybe a microcontroller monitoring temperature delta could work, or more dangerously, drilling a hole in the side of the chamber that would allow a miniscule amount of microwaves to escape into an RF detector. Alternatively, a camera can be set up to take a picture of what's inside every time it's turned on after a door is closed, or a logic circuit (perhaps with mains relays because there is no low voltage supply readily available) to only allow turning on when the door closes first, timer is set second.

Ohhh I though that the boss said "you can't go home early but that's okay because we'll all soon go home"