BUG: Small window of downtime when changing config variables or redeploying to Heroku

[tiblu]

Overview

There is a small window of downtime when Etherpad config variables are changed or a redeploy happens.

Thoughts

Citizen OS API had small downtime issues, we fixed them using the Preboot (https://devcenter.heroku.com/articles/preboot) feature of Heroku BUT I think this CANNOT be used for Etherpad as Preboot runs old and new instance in parallel, but EP does not support horizontal scaling due to not being stateless - it keeps patches in memory and flushes them to DB from time-to-time.

TODO

• Find out what causes the downtime and fix it, IF possible