Upon acceptance, some journals offer open access through an optional fee. This model differs from the usual open access model in which all authors pay a fee and all accepted papers are accessible without restriction. Addendum: the latter is referred to as simply open access, while the former is referred to as hybrid open access.
My question is as follows: I publish (most of) my papers on the arXiv, and as such, the preprints are picked up by search engines. When the corresponding camera-ready journal version appears, anyone with an internet connection can still access my arXiv paper, and hence I always opt not to pay the optional fees to make my paper open access.
In fact, most respected researchers in mathematics/computer science/physics/etc. also publish to the arXiv. As such, I am confused as to why this model of open access exists. What demand is it satisfying? Is there some advantage to paying these optional fees that I am not seeing?
P.S. -- The journal I have in mind currently is a SIAM journal. Their open access policies are listed here. Their open access fee is $2,500 USD (this is not an unusual number for publishers following this approach). This (at least to me) is a substantial amount of money per paper.