We were having scaling issues with one particular customer, so I looked into the various solutions. They all looked pretty cool, but I didn't like that the developer communities were small compared to MySQL and PostgreSQL.
So I ended up sticking with MySQL and building a simple sharding mechanism that allowed us to move big customers to their own server if necessary. It worked great.
While I was working on that, MixPanel published a long, highly detailed blog post about why they had moved to MongoDB. That led me to have second thoughts. "If these guys are doing it, maybe we should too?" But I was pretty far down the sharding development path and couldn't justify making such a wholesale change based on one blog post.
Fast forward about nine months, and MixPanel published a new blog post about why they had moved off MongoDB. Vindication!
Ever since then I've had conservative disposition when it comes to deploying new technologies.