Your first point sort of misses the idea that multi-tenant SaaS applications could and have shut down their service overnight (see Smyte https://news.ycombinator.com/item?id=17371074). At least with HipChat On-prem the customers who are running that version can migrate on their schedule instead of just having the service turned off.

You can always make the argument that for this reason we should only write our own internal tools and use OSS. However, it is just as easy for an OSS project to be abandoned or neglected. Often times enterprise contracts will have a source code escrow clause that provide the enterprise with the source code in the event of a shut down (the same capabilities as a fork & take over maintenance anyway).

We don't actually know the plan for HipChat On-prem yet (do we?). Slack and Atlassian are both known for being customer centric so they will probably follow what Facebook did with Parse. Open source the entire thing, sunset the operations and development over the course of a year or two with minimal security updates/patches and no feature dev. This could be what the investment from Atlassian was tagged for... to make sure their customers are taken care of while most migrate to Slack.