Whether you have multiple possible entries into the loop is irrelevant, what is important is whether you have possible exits.
As to your second point, does that mean it is ethical to run a simulation of someone being tortured as long as that simulation has already been run sometime in the past?
Possible exits could emerge from whatever the loop gets embedded in. (see 3 below)
Assuming a Tegmarkian multiverse, if it is mathematically possible to describe an environment with someone being tortured, in a sense it "is happened". Whether or not a simulation which happens to have someone being tortured is ethical to compute is hard to judge. I'm currently basing my hypothetical utility function on the following guidelines:
If your universe is casually necessary to describe theirs, you are probably responsible for the moral consequences in the
via.
Sounds like sour grapes. I'd heard of people holding such sentiments; this is the first time I've actually seen them expressed myself.