An essence of being an SAS is to perceive lawlike behavior in the universe it inhabits. Laws are essentially ways an SAS organizes what it observes about the world around it. In this way an SAS is able to learn from its surroundings, generalize event patterns into laws and constants that characterize its universe, and predict the likelihood of future events. An SAS with finite knowledge will allow for rare, apparent violations of laws, but assumes that an explanation exists for that apparent violation, which fits in with the laws it has knowledge of.
So it is no surprise, that, for example, a dragon event is never observed. If some SAS reported the sighting of a dragon event, its fellow SAS's would pass it off as a hallucination. That would be the 'natural' (i.e., fits in with the known laws) explanation. Similarly, if a group of SASs reported the dragon event, it would be passed off as a group hallucination, or possibly, a hoax. If there is an 'objectively' confirmed string of dragon sightings by SAS scientists, either the known laws would be updated while maintaining their simplicity (if that were possible) or else there would be great effort by the SASs to find a 'natural' explanation (e.g., the SAS's passed through a wormhole into another universe). Similarly, the Strong Anthropic Principle is really a statement that any SAS will find itself will observe itself in a universe that allows it to exist. The SAS should be aware which conditions that allow SAS's to exist and conditions which do not allow SAS's to exist. The appearance of fine-tuning really becomes rather subjective - how narrow should the range of values of allowed parameters should be so that there appears to be fine-tuning? Interpreting MWI is very difficult - I prefer to think of "many histories" myself. There are countless universe-histories which follow the same laws of physics that we have come to know and love. I am not sure if it is possible to talk about the actual probability of a particular history of our universe having occurred, or the probabilities of possible futures of our universe. Given the history of the universe up to now, the algorithmic complexity will presumably be the same for each member of a large set of possible consistent futures. The number of members of this set is unknown. For another set of equally less consistent futures, the algorithmic complexity is probably higher, but again the number of members of this set is unknown (though we expect it to be smaller).