To be fair I think there's a certain amount of momentum and economy of scale when you reach enterprise levels of anything. Incrementally it's quite cheap to toss one more of widget X on the pile when you have a room full of widget X experts.
If you decide widget Y is now the way to go, you get to answer all the questions already answered for widget X: Who supports it? How? How to do we back it up and restore it? Who can tune queries? How does this differ from our build patterns for servers that host X? Can the two live on the same machine or are there potential conflicts? Etc.
Don't get me wrong - with proper planning you can educate the X guys on how to support Y and smoothly add it or even transition to it. But for large-scale systems it requires a great deal of planning and forethought if you want to do it without any bobbles.
If you decide widget Y is now the way to go, you get to answer all the questions already answered for widget X: Who supports it? How? How to do we back it up and restore it? Who can tune queries? How does this differ from our build patterns for servers that host X? Can the two live on the same machine or are there potential conflicts? Etc.
Don't get me wrong - with proper planning you can educate the X guys on how to support Y and smoothly add it or even transition to it. But for large-scale systems it requires a great deal of planning and forethought if you want to do it without any bobbles.