Ask HN: Hard to Measure Design Decisions?

Some major design decisions like type safety , rdbms and monorepo have seen trends transition in favor , against and back again.

What are some good examples of “big” design decisions that are hard to measure conclusively?

- type-safety (eg perl/python/php were first made popular by lacking it, and now we are returning to it)

- rdbms vs nosql

- monorepo vs microservices

Others ?

This isn’t about advocating . Let’s think about what other concepts are hard to reach a conclusion about and why measuring success can be difficult .