Best practice is to host the status page on completely separate infrastructure. This adds a lot of operational overhead often seen as disproportionately low benefit relative to the ongoing effort to maintain, until rare events like this actually happen.