Read Anton Golov's answer to if our universe was a simulation what could a bug look like?
To crash it you need a bug in the simulation to exploit; and very few such bugs are actually going to crash a simulation; most will appear to be some sort of anomalous or inconsistent physics.
For example, in simulating movement of objects, I once had a bug in which my simulation let one type of object pass right through other objects, instead of colliding with them.
A simulation could fail to translate physical position under movement: To us that would look like a moving object suddenly jumped in space or just disappeared. Or perhaps the bug looked up the wrong physical characteristics, and an object (as small as an atom, big as a world) suddenly turned to solid uranium or hydrogen.
Most bugs do not crash programs; they just make the program do something unexpected. Inside the simulation, this would look like a new law of physics, or just something we never can explain. (Think of mysteries of the universe we have now, like Dark Matter, Dark Energy, the lack of monopoles, and several other anomalies in physics). It would be something that does not seem to add up or comply with the physical laws we have discovered about how the world and universe and all its systems work (including biological, chemical, physics and particle physics).
To Crash It, you'd need a runaway bug; and from inside the system this would be either unnoticeable (like time comes to a stop) or some kind of exponentially growing "infection" of the universe that destroys everything (e.g. a black hole that is unexplainably growing fast and will swallow everything.
The manifestation of a crash of a simulation is always a halt of the simulation; the program stops running. Now the computer I am running on may be processing instructions at full tilt; but if I am in an infinite loop the simulation isn't going anywhere: It is halted.
Most crashes are a result of bugs that accidentally over-consume resources, like computer memory or storage space, so the machine doing the computations for the simulation (which obviously is not IN the simulation) does not have the resources to continue. Or, giving such a machine orders it cannot complete; like to decide on a course of action using a variable that has accidentally been left 'undefined' in the code.
You might not ever see it.
IRL, many simulations take very long for us to run (Last year I read Military researchers required about a week of calendar time to accurately simulate air flow for thirty seconds worth of flight for a new kind of multi-rotor drone supply platform).
Because of that, many such simulations take check-points, where they store a 'snapshot' of the entire system every so often (like every 15 minutes), both to help them track down bugs and crashes, and if they are confident the problem did not affect the simulation up to some previous checkpoint, they will restart the simulation from that checkpoint.
From inside the simulation, we would not ever see the crash:
Suppose a simulation crash in our 2015 leads to the destruction of the universe. Our simulation programmers, trace this back to a bug that first manifested itself in our 1987: When Fleischmann and Pons first recorded cold fusion. But unlike our history, that turned out to be real, due to a bug in the simulation's treatment of interactions between palladium and deuterium. Once this interaction starts producing free energy, a runaway reaction causes a black hole that consumes the universe. Bummer.
The simulation programmers investigate this, and being clever, they figure out a way that Fleischmann and Pons might have seen what they saw, but with their results being due to some sort of intermittent equipment problem (and their experiments were intermittently 'successful' and not) localized to their lab.
They have a checkpoint from 1972, with a slight modification they can introduce this equipment flaw, and also correct the interaction bug. They restore the universe to that state: Things work out ever so slightly differently. Palladium and Deuterium interactions produce zero free energy; the bug is corrected. The equipment flaw trick Fleischmann and Pons, but their experiments cannot be replicated and they are discredited. Nothing at all happens in our 2015, and the simulation continues: But the simulation programmers have saved 13.7 billion of our years in simulation time, and perhaps that makes a difference to them.
