I would say that, as in nature, it is a fact of life that all systems need a certain amount of downtime.
Running your own infrastructure means that you can schedule downtime. If there is an important event such as a visit by the Queen, you make sure that there is a minimal chance of downtime at the most critical times. No updates, no testing, no development work, etc. Note the use of the word minimal there. Of course you can't factor in a squirrel chewing through a mains cable which is an event that is out of your control.
Using cloud infrastructure means that you've not only got the squirrel. A hell of a lot more now depends entirely on events out of your control. Earlier this week, as with any week, accountants had prominent events far more regular than the Queen visiting that would have required that conversation between Accounts and IT: "no dodgy work whilst raising invoices, end of month, payroll runs, VAT reconciliations, etc.", all of which would have got the Queen treatment to a greater or lesser extent. But this happened...
It's not as if you can ring Xero and say "look I'm doing my end of financial year-end today, would you mind frightfully not doing any work that can impinge on this please?"
You might say "Cloud? Downtime?" No such thing. Viewed externally, that's how it is perceived, but whatever they say, cloud systems need the same kind of maintenance we all need, it's just that it's done on a rolling basis, and we are not generally aware it happens... until it fails.
The point is that "maintenance" is a risky business because often there are dependancies involved: in the cloud, moving from one network to another to perform maintenance on the one going off-line means ensuring that the data is up to date on the network going on-line, and that all peering is pointing to the new one ahead of the switch. That transition is a work of art because it has to cope with a very large combination of what-if's, each of which aborts the transition if it fails. The idea of "routine" has moved up several notches, but is still vulnerable. With your own infrastructure you, as being responsible for any cock-ups, just don't put yourself in a position of risk. Ok there are those that do take risks, but you are engaging competent IT people right?
With the cloud, the competency of the people you employ to run your IT infrastructure is replaced by those that know how to manage cloud resources with the best outcomes. Those resources are a lot more complex than those pertinent to in-house IT, and just as expensive. Don't believe that the cloud means you can totally get rid of IT-savvy people. Cloud choices are often couched in administrative or cost terms, but there is an important technical element to them too, which needs addressing by a suitably knowledgeable person. How likely is that person to understand how a particularly cloud system works? It's not like the basics of how TCP/IP works, or how data on a Hard Drive is stored. It is at a much more abstract level than anything most of us have been in IT for years would be used to.