Previously, a big part of the difficulty with deploying MediaWiki in a timely fashion was that many deployments require an all or nothing approach. We didn't have a systematic way of deploying different versions of MediaWiki to different production wikis (i.e. previously, Hebrew Wikisource and English Wikipedia couldn't run on different versions of MediaWiki).
The "Heterogenous deploy" project gave us the ability to stage new versions of our software to production environment, and then deploy in waves to different parts of the cluster. This allows us to deploy in such a way that we can receive and address reports from the first wikis we deploy to before moving on to other wikis that may have different problems to report.