Investigate support for zero downtime deployment

[pablitar]

Currently updating catalysts ends up with around a minute of downtime. This is not too bad, but we could do better.

We should investigate ways to support zero downtime deployment

[menduz]

I've been thinking about this idea for a while: updating comms, lambdas and content independently instead of killing all of the images and starting up again