Scaling from 0 to 1M users: what architecture changes were necessary?

i'm curious to hear real-world stories about scaling an architecture from 0 to 1 million users. what were the major architectural pivots or bottlenecks you hit along the way? for us, the first big hurdle was outgrowing a single database instance and having to introduce read replicas and then sharding. caching became absolutely critical around the 100k user mark. i remember a point where our monolithic service was just not cutting it, forcing us to move towards microservices to scale different parts independently. what were the crucial decisions you had to revisit? which 'best practices' turned out to be premature optimizations, and which ones saved you down the line? i'm looking for candid insights into the practical journey, not just theoretical concepts.