This canary is a good thought. The problem the article highlights is that people don’t practice updates enough and assume someone else or something is handling it. You only get better at it the more often it happens which is partly why long expirations are not ideal. This is what the article is highlighting as the main issue.

It’s not a good thought. Run a single client (uptime kuma) and ask it to alert you on expiration proximity. I.e. implement proper monitoring and alerting. No need to randomly degrade your users’ experience and hope they’ll notify you instead of shrugging and going to a site that doesn’t throw made-up http errors at them randomly.