This is actually a generic problem in public acquisition - projects are tightly funded to meet specific customer requirements and cannot themselves resolve enterprise-level problems. Building for re-use tends to add complexity / cost and a hard-nosed PM will not easily be pursued to solve someone else's problem. It seems to require top-down commitment of intent and resources - and a big stick - to make individual projects do the right thing.
Obviously there can be situations where common approaches are developed and used but this seems to be the exception rather than the norm.
(source - I've spent 10 years trying to work this issue in a UK public acquisition context)
This is an issue in enterprise situations as well, I have 100s of servers which are supposed to be "managed" but each project / product manager just adds their own company card to amazon and builds their own platform. The only way this can be solved is if it gets escalated to VP / CTO levels and they force everyone to follow the "standards".
oddly enough most enterprises I know have servers "standards" for their own data centres and would dearly love to just fork cash to amazon to get them out of that hell
Yup. The heads of IT at enterprises are begging to just switch everything to AWS/GCP/etc., while all the mid level IT guys are starting to see the writing on the wall. Much of what they do will not exist anymore and they will need to retool themselves and/or find a new job.
This is quite true. At the end of the day, you've got to meet the requirements, and there are design choices you can make so that the solution for one customer is clean and elegant and yet completely unusable for another customer.
In any build vs. buy situation, "build" is driven by the desire to enlarge someone's empire, while "buy" is driven by personal relationships and other sales techniques.
The first maximizes "unique requirements" while the second minimizes them.
Between the two, I generally prefer "build", bad as it is, because it leads to fewer horror shows later.
Obviously there can be situations where common approaches are developed and used but this seems to be the exception rather than the norm.
(source - I've spent 10 years trying to work this issue in a UK public acquisition context)