How do you manage rare events in such a setup? For example, how to restore a database from backup is valuable knowledge that should be documented somewhere so y

How do you manage rare events in such a setup? For example, how to restore a database from backup is valuable knowledge that should be documented somewhere so you don't have to make it up as you go along. At the same time, in (almost) every place I've ever worked at this was not required very often. This seems like something chaos engineering and/or a proper ops team with regular disaster exercises would be able to pick up, but (like you mention) in most startups those seem to have very low organisational priority.