There’s a big difference between deploying 3 times per week to your development environment and by doing so constantly updating and exercising your deployment infrastructure so you know it will work when you’re ready to rollout to production (all of which makes perfect sense to me, I like what you’ve described here); and the extreme case described at IMVU, where programmers push changes directly to production after a successful automated regression test run, up to 50 times per day. At that extreme I don’t see how continuous deployment allows for sufficient controls and reviews for reliability, privacy, security and so on.

We push changes to our development and test environments several times per week (or even per day), but we run through an additional set of checks and reviews and tests before deploying a release to production every 2 to 3 weeks (although of course we can push out a hot fix or small changes in the same day if required). And of course we use “dark launching” and feature controls and other techniques to minimize the risk of change to our customers. We’ve learned to do this over a few years and it works for us. I can see how some teams can run faster, and why some need to take more time before rolling out to production, there’s a balance that needs to be attained. But I can’t see how releasing to production continuously several times per day can be sustained – there’s not enough time to do a proper job on everything.