There is a MNer you would be interested to talk to. She does meta-analyses and has done a few on homeopathy. Maybe searching homeopathy & meta-analysis on MN would lead you to her. She explains the nature of statistics and the conclusions of these meta-analyses better than I ever could.
"prob of a god: very low x negative outcome for not believing: eternity in hell"
You can't know "probability of God". You can know probability of rolling a 3, because there are six sides to the die and only one 3 = 1/6. You can know probability of drawing an Ace because there are 4 in a deck of 52 = 4/52.
You could only know the probability that our universe was created by a God if you knew many universes and the number of them that were created vs others that spontaneously came into being.
Also, I'm also totally ok with the possibility that there may be a God, but all evidence I see and my rational mind says that he must be a watchmaker God rather than one who interferes and judges. Therefore, I don't see the outcome as "eternity in hell" at all.
"what I see as a single obsession with one potential trigger of autism"
We would be "obsessed" with other triggers as well, if those had large numbers of parents very convincingly saying that they triggered DCs' regression.
"incidence of this mutation was 1 in a billion i.e. there are 7 people in the world at risk, by your reckoning, you STILL wouldn't vaccinate as the outcome would be too bad"
It would depend on the benefit. If the vaccine in question is to protect against rubella, I wouldn't do it, because it's silly to vaccinate a baby for what is an adult problem (in pregnancy). If we are talking about HPV vaccine, I would do it.