You can consider performance bottlenecks as a kind of Technical Debt, so include this issues in your tracking. Distributed systems have their own idiosyncrasy manly due to latency in communications and transactional operations. If you want to think in the near future performance, the best thing you can do is: - test the system in a big scale - collect bottlenecks (some times you need and special toolbox and people with experience to find the root cause) - solve the issues One thing I do in my work, is to locate small changes in code that have big performance impact. You have to pay pay attention to details. Data structures and algorithms are essential to improve the performance, and sometimes they can be changed or improved easily.

When it comes to avoiding performance bottlenecks, the key isn't trying to future-proof everything but designing smartly to isolate potential problem areas. Focus on minimizing the impact of code "hot spots" and be mindful of communication types. In-process calls are efficient, while out-of-process and cross-network communication can add significant latency compared to in-process interactions. Keep slower components off the critical path; add metrics using tools like Application Insights or Prometheus to catch issues early. Just remember, metrics can add overhead, so balance their use wisely. The goal? Catch and fix issues when they arise, not over-engineer to avoid them.