I'm guessing you have a typo above, because you have 10. and 100. listed.

If you get this sometimes, could the cluster be failing to the other node? Clusters do no get 100% uptime. When there is an issue, the failure requires the server to restart on the other node, which can be a 30s or so delay.

Have you checked this? This is the most likely cause.

Note, your code should a retry rather than erroring right away to the user.