It would be nice to have a proper safety margin of redundant machines - say, one hot spare for every core machine, and one for every two apaches/squids - before deciding that "money for hardware really isn't the issue". I had the impression that some of the recent delays and outages would have been less troublesome (and less noticeable), had there been more redundancy of key components.
The $50k in available funds as of Dec 31 was a bit less than two months' operating costs; not as much as it might seem.