There’s a reality somewhere. A “physical” universe in which a computer can be created. With enough time, a computer can be created in this “top level” universe that is capable of simulating one universe. Then, assuming computational power and efficiency can be improved, multiple universes can be simulated at one time. You can also achieve better simulation fidelity by slowing down the simulated universe, like 1 second of simulation time taking 1 minute of top level universe time.
If we stop there, the probability that we’re living in a simulated universe instead of the top level universe is already pretty high (or inversely the chance that we’re living in the top level universe is pretty low).
Now, if the computers are powerful enough, the simulated universes can probably have computers in them, and those computers will eventually be able to simulate universes, too. Probably in about the same ratio.
So it’s not simulations ALL the way up, but if those postulates are rational, the chance that any randomly chosen universe in the set of all universes is the top level universe becomes vanishingly small, but non zero.
I think it’s definitely the case that if the top level simulation(s) stop, all the downstream ones would, too.