So I've been bitching to one of our devs about how his updates to a replicated table suck. Today I went looking. What the.... it issues a delete, then inserts the row? Why are you doing it that way? Well, he's not. SQL is. : - \
Turns out, if you have a unique index or constraint (as opposed to the default PK index, I suppose...), then it deletes then inserts the row. Which sucks when you have a massive wide table with a ton of records.
Live and learn. Hopefully this helps somebody else.