Well, you're restricting your imagination somewhat. The clock inside a simulation is self-referencing. It doesn't half to self-heal. The operator could run a single cycle, see what happens, then make adjustments before continuing. It doesn't matter how long it takes to calculate or process the physics of the universe, since the referential time-frame of the subjects of the universe is part of the simulation itself. One second in the simulation could take 10 seconds, 10 years, or a hundred million years to process in the "outside world", but it wouldn't make any difference to the operation within the simulation.
A lot of commenters have posted arguments like the scale being impossible (why? our simulated universe could be 10^200 times smaller than the universe one level above us, there's no reason to think it'd contain the same amount of information), or that things like quantum physics would break the bounds of the computer running the simulation (sincere lack of imagination - if it's a simulation, these quantum physics don't "exist", the simulation of them is part of a running program), and other such arguments which keep forgetting that they're talking about a simulation, not something physically real.
It's basically comes down to this - if we postulate that given enough time (billions of years? trillions of years? pick a number) and sufficiently advanced technology, we could eventually simulate the entire physics of our universe (the nature of the physics) and accurately simulate a universe (even if much smaller than our own) right down to every aspect of the physics, then it immediately becomes more likely that we're in such a simulation than not.
Why? Because if we were able to have an accurate simulation of the physics, then it would be possible for the inhabitants of our simulated universe to also do the same, on a smaller scale. And so on and so forth. There's an analogy in computers in that we can write an emulator which accurately emulates in software the hardware of a computer. Then, in that emulated computer, we can write another emulator which accurately emulates in software the "hardware" of the simulated computer, etc etc. The limit of this recursion is only the processing power of the initial "seed" computer. However, each computer in the chain acts as if it is the seed computer, and its processing power is the first step.
Likewise, in a universe simulation, the inhabitants of each universe would think they're the top level "seed" universe. As it's a simulation, there's no way they could ever escape the bounds of the simulation as they don't "exist" outside of it. Thats what makes this simply a thought exercise, because even if simulated, it's still our "real", as you mention at the end of your post.
But, if we accept the initial premise that we'd eventually be able to produce such a universal physics simulation, that makes it logical that we're in such a simulation, and by no means the "seed" universe. After all, if we could run one such simulation, why not multiple. And why would the inhabitants of those simulations once they too were able to do the same (given they have the same physics, it would be possible within the simulation), also not run multiple simulations? If we accept the initial premise, then we have to accept that there could be an uncountable number of simulated universes, only one seed universe, and no way for the inhabitants of any simulation to ever know if they're the seed or a simulation.
And it doesn't matter. If we're in such a simulation, we could never know. Even if the operator decided to program the simulation to be aware of itself, it would still only be operating within the bounds of the simulation. What we think of as self-awareness would just be part of the physics simulation running its course, meaning that although the inhabitants of the simulation would 'act' self aware to an observer, they're only doing so because of the accuracy of the physical simulation.
It's all a fascinating thought exercise, but has no real world applications. For us, the simulation and reality would be one and the same.