Most providers advertise 99.9% availability, yet the number says little about how quickly support answers at night or how often a node is overloaded. We now track response times ourselves over several weeks before writing anything.
For a blog with a few thousand visitors a month the difference is barely noticeable. Memory and a sensible cache matter far more than the disk type.
Update packages, switch SSH to keys, enable a firewall, set up automatic certificate renewal and keep a copy of your configuration somewhere else. Boring steps, but they prevent most bad days.