I don't think it makes sense to treat a simulation where you rapidly change into an unrecognizable mind differently from a simulation that halts, except insofar as the transition might be unpleasant. In either case, the experiences you value have a certain measure up to a certain point in time and then they don't.

I agree that dreams are still highly ordered despite being a bit less so than waking life. I think the experience of a human dreaming in an orderly universe is still most likely caused by a human dreaming in an orderly universe. But I don't think it would make sense to be scared of sleep even if it were otherwise.

The way I'm looking at this is that your goal should be to take actions that will increase the measure of experiences you value and decrease the measure of those you do not. This adds up to normality: Sleeping increases the measure of 'dreaming' experiences and the measure of 'waking up with enough sleep' experiences, and decreases the measure of 'being awake at night' and 'being unrested in the morning' experiences. That's the only effect it has on the measure flows of the multiverse.

I think the way you're trying to look at it in terms of a thread of experience is at best extremely difficult to follow and at worst flat-out wrong, and you'll be better served to decide which measure flows you want to cause (of those you can cause).