Why can tempdb become a performance bottleneck?

tempdb can become heavily used by temporary tables, large sorts, hash spills, row versioning, index operations, ETL, and other internal operations.

Potential problems include storage latency, insufficient capacity, excessive autogrowth, allocation contention, and queries that spill large amounts of data because of inadequate memory grants.

Troubleshooting should identify which workload is consuming tempdb rather than assuming that adding more files will solve every tempdb problem.