Very interesting thread, although I suspect some of the disconnect here is what constitutes an outage acknowledged on a status page vs what is "just" a higher than typical error rate that hasn't yet violated an internal SLO.
It's never a good sign for any engineering team when customers detect an outage before they do. If this is hours before: that is embarrassing.
Render (a customer of Google Cloud) detected the outage ~3.5 hours before Google Cloud acknowledged it.
Does not instill confidence.