But we still have use cases where we don't use these DB - partly for cost reasons and partly because we want different functionality. Berkeley DB ldap for accounts, ETL and analytics moving into Hadoop, some apps use Mongo and even other smaller apps use MySQL, and those are just the projects I know about from my view. We still pay for support contracts on all these projects but obviously costs are significantly lower.
Even in a company that doesn't mind paying large amounts for enterprise database, I don't think that that reaching for one of these major commerical DB is always the right solution. Not all projects need the support you're paying for from them, and beyond that for some useage patterns I don't think they're even the right solution to reach for, regardless the cost.
> It says a lot when a Big Company dumps a Big Database vendor and moves to Oracle to save money.
I do not know which Big Company you are referring to? But also, never trust a headline which says "Big Company does X" to tell the whole story. It's very often part of a larger deal, a big discount given to them specifically for that headline, or some soft "nice business you have here, wouldn't it be awful if some patent litigation ruined it? That can be avoided if you buy our products" talk.
Everyone is assuming they will need to grow quickly. The costs of the systems you list become prohibitively expensive when you grow.
If e.g. Google or Facebook had to pay for an Oracle or SQL Server for their storage (in case those DBs could actually deliver the performance Google needs), or, for that matter - Windows Server licenses), they would have had to raise a lot more money and dump it into Oracle/MS. It is this that doesn't sit well with SV culture.