The timeout for the requests issued by the uptime monitor was set a little too low – the uptime monitor mistakenly believed the service was not responding (it was, but slowly)