But the article is about a personal website. Websites run over HTTP, which is not designed for anything unreliable.
It's interesting to think about a web that was designed to have lots of intermediary caching peers along the way as part of the protocol, but that's not what we have.
The notion of mobile-device-as-origin-server which can provide either content in a local context (on a LAN, localhost-only), or globally via cache-and-redundancy / cache-and-forward networks has definite and positive use-cases.
The sending client is supposed to retry if the receiving server is down.
But it's still built with the expectation that the recipient host will usually be available, and that unavailability is a transient condition -- a failure to contact the recipient SMTP server is a noisy failure, and, after the first few retries, most servers back off to one attempt every 8 hours. Mail servers which are only sporadically available would require some fairly substantial rearchitecture.
The noisy failure is probably not a part of the standard, more likely a recent trend.