I broke a few things in a kind cluster to see what kwatch would catch: crash loops, a bad image tag, OOM kills, and a database going down. It caught all of them. My favorite part was seeing seven failing pods grouped into one alert during the database outage. When several things break at once, having fewer messages to sort through is helpful.
Breaking it on purpose in kind is the best way to see what it does, thanks for writing it up. The image and memory limit you wanted are in Pro, along with what changed and the command to fix it. Pro's free while it's in early access, you can turn it on from the License menu in the installer. For your DB test, Pro should say the database is why those seven pods crashed, not just group them. Would love to hear how it goes if you try it.