Regarding the last paragraph, I was once in a meeting at a very large company in which an AVP asked me and several other experienced engineers: "Ignore the technical details; is there any reason we shouldn't purchase this large suite of Oracle products?" That was the litmus test, so yes, it's purely a safety play for many.
So for most companies UX is more important than how messy your backing store is. And yes, delivering good UX requires quick iterations, and that probably means a good operational support for experimenting with various data access patterns, but - again - SQL is enough. Just throw more Hadoop at it, if not. (For specialized data stores. Or just more hardware and even more 12c instances and use that as your key-value store. Doesn't matter. Wrap it and spend more time on UX and features.)
Yes, disrupting them with better technical skills is possible, but don't forget that in the end you have to provide value, and rarely a "contains only F/OSS" sticker is enough.
In addition, their portfolio is so vast that any large business will likely have something else that can be leveraged in any discussion ("don't like our db pricing? Let me cut the renewal price of this other product you bought last year. Or what if I throw in a few licenses of that other product you were already evaluating?"). This also allows one good product to cover for the shortcomings of others ("ok, this project failed horribly, but that other thing last year was hot stuff! You got to give us another chance!").
Of course, they already are on the list of approved vendors, so there is no bureaucracy or exposure for whoever signs the purchase order. CYA rules.
If you want to get really technical, however, most Oracle products will just "tick all boxes". They might not be the best option, the fastest one or the easiest to maintain (and they are never the cheapest), but they're likely the most feature-complete. For example, the db: want multiple active-active instances? can do. Active-passive? can do. Multiple names for same instance? can do. Multiple instances on same box? can do. Cross-platform? can do. Working without DNS? can do. Objects? can do. XML? Can do. Your company has weird storage requirements? No prob. Maintenance tools? More than you will ever use. And so on and so forth. On paper (which is where executives live), it's hard to justify anything else; problems only come up when you actually try to use the beast.