In my experience, multi-hour control-plane outages are very rare, and I've only had a single data-plane outage (in this case IIRC it was a GKE networking issue). Worst case I see is once or twice a year I'll get a node-level issue, of the level of severity where a single node will become unhealthy and the healthchecks don't catch it; most common side-effect is this can block pods from starting. These issues never translate into actual downtime, because your workloads are (if you're doing it right) spread across multiple nodes.
I wouldn't be surprised if EKS is bad, they are three years behind GKE (GA released in 2018 vs 2015). EKS is a young service in the grand scheme of things, GKE is approaching "boring technology" at 5 yrs old.