How?
An example of a test that's phenomenally complicated is, roughly speaking, to zoom in on physical reality until the resolution starts to drop https://en.wikipedia.org/wiki/Simulation_hypothesis#Testing_...
An example that's like magic is to communicate with someone outside of the simulation. Either someone in an adjacent simulation, or someone outside of the simulating system itself.
Basically, a simulation has to have rules and constraints. If we can get a fingernail into a crack we can expand it. Just keep pushing the limits until they break. We could even start the simulation version of SETI. That's a project that's testing the hypothesis that there are ETs out there we can communicate with. We could start a project like that based on the hypothesis there are Simulators outside of our simulation we can communicate with.
Unless we reduce the simulation hypothesis to "Beings of similar intelligence level as us running similar computer simulations that we do in the late 20th and early 21st century", then it is testable.