edit: I saw the same on a fleet of thousands of JVMs which hung on 100% CPU after 248 days very consistently. Closest thing to an explanation I ever got was perhaps it is storing uptime in hundredths of a second (why not ms???) in signed 32 bit integers, see: https://ma.ttias.be/248-days/ In the end we solved it by restarting with a cronjob between 2am and 4am after 247 days...