hmm classics would be

• single or multi-tenant system ? -> if multi think about splitting them into separate databases
• Roles and Permissions this should be thought of BEFORE not afterwards
• Auditing -> we audit everything per default, but we only activated auditing for the crucial features in production
• Errors -> there are two types of errors -> errors for the dev / ops -> clear, structured "where, what, why" the customer how ever should not get system information
• CI / CD ... start with tests, I don't care if it's TDD, BDD or whatever just test as much and as good as possible and start simple -> the bigger it gets the more complex changes are and the more things will explode in your face
• caching -> no caching till you hit the bottlenecks ! if you have a clean layer structure you can set new cache layers everywhere but it will increase the complexity.
• DB -> check what you need and how the data comes in and go from there ... you can scale with a lot of different databases so this is really about "what do I need" there is no silver bullet
• Technology is your choice you can scale with almost any language and tech.
• Try to use a log system like Logstash and collect metrics with graphana or so.

those are the first things that come to mind