Since write conflicts are normal in multi-threaded workload that heavily updates same documents (ex: YCSB with latest or zipf distribution), the logging of it should be significantly lower.

Recommendation:

• LogLevel 0: log nothing (updates with a lot of writeConflicts will be logged as slow ops if they exceed slowMS threshold)
• LogLevel 1: log update conflicts > 100 attempts
• LogLevel 2: log update conflicts > 10 attempts
• LogLevel 3+ : keep at current logging or log all.