I've come up with a model of how online discussions of perfectly reasonable topics go toxic. It's based on the proposition that how one behaves during a discussion depends both on what one's opinion is and on how deeply the opinion is felt. You can have a very high opinion of something but not hold it very deeply, or a low opinion you don't hold deeply, and so on.
There are other dimensions to one's opinion and behavior -- how much doubt one has of his opinion, for example, and with how much humor he hold his opinion -- but let me keep it two-dimensional for now.
So we have something like this to model attitudes toward some particular thing:
As you can see, there are three zones in this diagram. Someone who has a sufficiently high opinion of a thing, given how deeply he feels it, is a zealot. Someone with a sufficiently low opinion, relative to depth of feeling, is an anti-zealot. And the stronger someone feels, the less highly [or lowly] he has to regard the thing to be a[n anti-]zealot.
By its nature, zeal for a particular thing regards discussion of that particular thing the way a child regards a motel swimming pool: it jumps right in at the first opportunity and won't come out until it's forced to. When zeal and anti-zeal are both present, they ensure that, whatever else might force an end to the discussion, it won't be a lack of things to say. I'll put my first, not particularly original, rule of thumb this way:
When a zealot and an anti-zealot are involved in a discussion of the particular thing they are zealous about, the discussion goes toxic.
If I think of a topic of discussion as a set of related particular things -- that is, of things one can have higher or lower opinions of -- then I can propose this additional rule of thumb:
If a topic includes two particular things such that a zealot of one particular thing is likely to be an anti-zealot of the other particular thing, and if zealots are not too rare, then discussion of the topic is likely to go toxic.
Since zeal is so reactive, it doesn't have to be very concentrated in the general population for it to be expressed in a discussion. And once zeal is expressed, any anti-zeal in the general population will respond.
This, as I say, is just the nature of zeal. Stones, when released, fall; zealots, when capable, express zeal. Non-zealots are fully aware of this, and by their nature are averse to the toxins a discussion between zealots and anti-zealots can produce. So as the topic is revisited within a general population, the concentration of non-zealots involved is likely to decrease. This leads to the third rule of thumb:
Discussion of a topic is more likely to go toxic again if it has already gone toxic before.
Again, nothing revolutionary in Internet sociology. It might offer some insight into the toxicity of discussions on the topic of liturgy, but since discussion of discussion of the topic of liturgy has proved toxic, I'll leave that as an exercise to the reader.