Summary

_extractReadConcern runs ~120 lines of branching logic on every operation to determine which read concern to use.

If the most common case is to use the implicit default (local), we should have a fast path for it.
A fast-path early return for this common case eliminates a mutex-protected ReadThroughCache lookup, and nested conditionals.

Profiling YCSB 100 read shows 0.85% of CPU in this path, dominated by mutex contention on the cache.

Proposed Fix

Add a cheap atomic boolean and accessor in read_write_concern_defaults.h and a fast-path early return at top of _extractReadConcern:

See this patch for proposed change. 

Impact

This multipatch shows a 1.3% improvement in our 100 read YCSB workload.