checks fire off a lot of parallel work, so its possible you're causing contention on your connection pool or due to context switching (which are the two typical resource exhaustions when scaling past your limits)