The config default inactivity timeout has long been used as a "cover up" to fix an inexplicable connection leak that'd escape all the more intuitive different timer transaction and connection level configs ATS supports (https://docs.trafficserver.apache.org/en/latest/admin-guide/files/records.config.en.html#proxy-config-net-default-inactivity-timeout) While the idea was always to use this as a guard timer (catch all) for the odd orphaned connection and as a result the recommended default was 1 day (86400 sec), it has been noticed for a long time that, during each release upgrade, this would result in running out of connections after a few hours. A workaround has been to set this timer to much lower value (5-10 min) in most production environments. Investigating the root cause behind why this timer was even needed in the first place, it turns out the problem was due to some bugs in connection queue management, which have been fixed in master/9.x with the PR (https://github.com/apache/trafficserver/pull/5824). Given that the root cause has been addressed, we'd like to deprecate the guard timeout for 9.x, and eventually remove it. Let us know if there are any concerns/comments. Thanks, Sudheer
