I lost half a day because BeginScope silently did nothing in production: no error, no warning, just a flat stream of undifferentiated log entries. ILogger is a façade over a pipeline full of opt-in behaviour that looks enabled by default. Scopes, structured properties, minimum levels, exception chains, timestamps: all have failure modes that compile cleanly and fail quietly.
The code you create is a valuable legacy — my author bio. Not marketing copy. It’s the most expensive lesson from nearly two decades in production systems. This article explains what it actually means when you’ve lived with the consequences.
DateTime.UtcNow looks harmless. It is not. It is a hidden dependency you cannot control in tests, cannot reproduce in staging, and cannot freeze to catch the bugs that only surface at midnight or on the last day of the month. .NET 8 shipped TimeProvider to fix this. Two years on, most codebases still ignore it. Some planned to adopt it later. Later has not arrived.
AI code review tools are genuinely useful for catching syntax errors, obvious bugs, and common anti-patterns. They are also systematically unable to tell you that the feature you built was the wrong call, that the abstraction is off, that the naming reveals confused thinking, or that the correct review comment is “delete this.” Here is what AI reviews find, what they miss, and why human judgment still has no substitute.
Source generators are powerful. They are also running on every single build, blocking IntelliSense, breaking Hot Reload, and multiplying their cost across every target framework you support. Nobody mentions this in the getting started guides. Here is how to measure the damage, find the culprits, and decide when source generators are actually worth it.
