Seems like there's a pretty big difference between two groups duplicating the work of a handful of engineers, like in your query engine example, and billion dollar programs like in the article. My personal favorite example of the former were workflow systems; somewhere around 2009 Google had like 15 different ones. But obviously the reason was that nobody had foreseen such a need early enough, a lot groups ended up building their own solutions, and with so much fragmentation none of them could reach critical mass organically.

In terms of large scale and crippling internal competition, Nokia comes to mind with their utterly non-cooperating consumer vs. business mobile phone divisions back in the day.