Rate limiting and bulkhead isolation are often mentioned together but solve different problems. This deep-dive explains both patterns, shows where they fail without the other, and walks through Granit’s dual enforcement model — Redis-backed quotas for the time axis, concurrency limiters for the thread axis.
From accidental client evaluation to N+1 queries and tracked reads, these are the five EF Core anti-patterns that turn 50 ms endpoints into 5-second ones — and the exact fix for each.
ASP.NET Core dropped the synchronization context years ago. So why do every Granit library and the .NET runtime itself still write ConfigureAwait(false) on every await? Because libraries do not get to choose their host.
Two-tier caches are fast, but cross-pod invalidation is the part everyone gets wrong. Here is how the L1 + L2 + backplane pattern works, and how Granit wires it up with Redis.
String interpolation in log calls allocates on every invocation, even when the log level is disabled. Source-generated [LoggerMessage] eliminates that cost entirely and gives you structured logs for free.
new Regex(…, RegexOptions.Compiled) allocates at startup and hides patterns from the JIT. Source-generated regex is faster, safer, and enforced at build time.
