May 2026 - Proton: Delivered major shard-pruning optimizations for IN-subqueries and literal tuple-IN expressions, enabling early subquery materialization, reducing shard reads, and improving overall query latency on large datasets. Added new control setting (optimize_skip_unused_shards_with_subqueries) and on-demand subquery materialization to ensure pruning only occurs when safe. Hardened NULL handling to avoid incorrect pruning; extended support for empty-IN in streaming paths with safe fallback to all shards. Expanded testing with SQL regressions and gtest coverage; deduplicated literal-tuple RHS handling and improved explain coverage. Refactored internal architecture for detail namespace and helpers to improve maintainability. Result: faster, more predictable queries with lower resource usage and improved streaming stability.