> Twitter famously had a "fail whale" but it didn't stop the company from growing. If you have market demand (and I guess advertising) then you can get away with a sub-optimal product for a long time.

Agreed, but there's still an element of survivorship bias there. Plenty of companies failed as they couldn't keep up with their scaling requirements and pushed the "getting away with a sub-optimal product" for too long a time.

Do you have some good examples?

This touches the toupet fallacy: "I never saw a large company fail to grow large because of deferred scaling"

Friendster might fit though: https://highscalability.com/friendster-lost-lead-because-of-...

I agree. Go fast with a suboptimal architecture. If success arise, throw away version 1 and rebuild from scratch. Often is more effettive.

Reddit is still around.