You would want to enable Debug Logging Level 3 on the Sentry(s) and see if you can identify why they are being marked dead in the Sentry log (log/tomcat/sentry.log). This is common is the response from the EAS back-end is taking too long and a timeout occurs or if the server responds with too many 503 errors. If the back-end is O365, you should disable Health Checks. If you’re using a hybrid of O365 and On-Premise, you should split your user-base into appropriate groups to make sure that each Sentry is not proxying for both O365 and the On-Prem solution.