performance-reviewer

To find N+1 database queries during code review, look for database queries executing inside loops that should be replaced with a single batched query or an eager load. This pattern causes sequential round-trips and bottlenecks runtime performance.

Unbounded memory growth is caused by loading entire tables into memory without pagination or streaming, caches lacking eviction policies, or string concatenation in loops. These hot-path allocations expand runtime memory indefinitely and stall the system.

To detect blocking I/O in async contexts, review code paths for synchronous file reads, blocking HTTP calls, or CPU-intensive computation on event loop threads. These operations stall other concurrent requests and severely degrade runtime performance.

Missing pagination causes out of memory errors because endpoints return all results without limit, offset, or cursor streaming. When consumers handle the full unbounded result set, the application experiences rapid memory growth and crashes on large data.

The best way to find hot-path allocations is analyzing loops or per-request paths for object creation, regex compilation, or expensive computation. These should be hoisted, memoized, or pre-computed to prevent repeated runtime overhead and scalability bottlenecks.

Runtime performance review detects scalability issues in caching layers by tracing unbounded memory growth and missing eviction policies. It surfaces bottlenecks in code paths with database queries, loop-heavy transforms, and I/O-bound operations.