Scaling Azure Functions & PaaS - Concurrency, Async, Messaging. Insights from Python Log Analysis

Nice write-up!

Interesting to learn that the Azure SDK for python is generally not thread safe, and the one for .NET usually (always?) is, as explained here for the service bus client: https://learn.microsoft.com/en-us/dotnet/api/overview/azure/messaging.servicebus-readme?view=azure-dotnet#thread-safety

Some remarks:

1. About the "Reuse credentials to avoid re-authenticating": auto-refreshing an expired token only works when you're staying in context of the Azure SDK, so just beware of that. At least this is the case for .NET, i assume that's also for python: https://learn.microsoft.com/en-us/dotnet/azure/sdk/authentication/best-practices?tabs=aspdotnet#understand-when-token-lifetime-and-caching-logic-is-needed
2. About Robust Logging: What about Function Apps that might not finish gracefully in rare cases, that can happen. How robust is your logging then? Any mechanisms you use for that? It seems like building for idempotency might always be mandatory.
3. You mention "after the nightly run": that sounds awfully close to batch processing, while at the same time you are making a strong point for event-driven processing. Are you doing some batch processing and is it unavoidable or maybe even preferrable in your case? 
4. And finally, a nitpick; you mention Azure AD instead of Entra ID 😁