The failure mode that costs us most is not the one where an endpoint disappears. We run a crawler over dozens of job boards every six hours for an iGaming index of about 6,700 postings, and the expensive breakages are the quiet ones: the response still returns 200, the selector still matches something, and we get forty jobs instead of four hundred, or the same four hundred with the salary field silently empty. Nothing throws, CI stays green, and we only catch it when a weekly number looks off. Hard breaks we find within the hour. The other half of what we depend on has no release to watch at all, like an undocumented JSON endpoint some frontend calls or plain HTML that gets rewritten without notice. Does Since treat those as dependencies as well, or does it need a versioned surface with a changelog before it can tell me something moved?
Since can already watch publicly readable pages without a version or changelog. But it doesn’t currently validate your crawler’s output, so I wouldn’t promise it would catch 400 jobs dropping to 40 or salaries quietly becoming empty.
Undocumented JSON response contracts and those data quality checks are gaps today. I’ve added them to the roadmap: monitoring JSON and HTML structure, plus per-source counts and field completeness. The important part of your example is catching incomplete results even when the request succeeds and the selector still matches.