There's a difference between being adverse to failing at some task and not wanting your venture to crash.

For vital infrastructure, it's more important that those "failures" be more in the way of the dba or sa needing to look up how to make something work properly, than actually fry the server instance and wipe your data.