March 2026 monthly summary for pmndrs/jotai: Delivered a significant enhancement to the benchmark suite for store operations performance evaluation. Replaced three simple benchmarks with eight comprehensive scenarios to broaden coverage, including atom-creation throughput, batch read/write on mounted atoms, deep derived-chain propagation, 1→N recompute performance, diamond dependency updates, subscribe/unsubscribe churn plus subscribe-write scaling, cached derived reads, and partial atom selection. Implemented a run-all script and a bench command to execute all benchmarks end-to-end, and added a bench script to package.json for convenient usage via pnpm bench. This work provides richer, scalable performance signals from 10^2 to 10^6 store size, enabling more accurate optimizations and capacity planning. The changes improve benchmarking reliability, reproducibility, and cross-team value for performance-focused decision making.