I know everyone is all about running huge clusters behind the scenes, but most people simply don't need that, and being able to start small, and buy bigger would be a nice option.
For example, a simple node application using leveldb(levelup) for the storage interface could be very effective as a backend for development/testing. From there, you provide an API compatible interface. Open-source that version, with the disclosure/understanding that the Azure hosted version is much more robust.
I think you'd see a lot more buy in from the open-source community... even more so if you accepted PRs to make the open version more robust.
If you are at a point where you are considering the likes of ES/SOLR/Lucene etc, you are likely ready to make the jump to self-host in the cloud, or use a SaaS provider. Where people get a bit concerned is in the lock in. I know why Azure would want to present that, but I think it's a bad idea without an open implementation that allows for self-hosting for development and on the small scale.
Right now, the company I work for is hosting in Azure. I recently switched from a couple RabbitMQ queues to Azure Queues. It works fine, as was a really simple replacement for a flow of near real-time but temporary data keys. I would be much more open to using a hosted MongoDB from MS than DocumentDB, in much the same way I'm happy to see you guys embracing Redis for a cache system.