Add a failOnError option to status-polling on the deployment page. The timer is still started when entering the page so that the status of running clusters gets displayed, but if the connection to the mgmd fails  we cancel timer and start it again when the mgmd is started. The timer is also cancelled when the cluster (mgmd) is stopped. (It is also cancelled when leaving the deployment page, as before). This will remove many bogus "connection refused" messages  which currently make the log hard to read.