qec-kiln paper-v1 artifacts: cloud-distributed Sinter for QEC Monte Carlo sampling

Raw result data and reproducer pointers for the arXiv paper "qec-kiln: Cloud-distributed Sinter for scalable QEC Monte Carlo sampling" (B. Irish, 2026). qec-kiln is a ~300-line distribution layer that partitions Sinter collect jobs across cloud spot instances orchestrated by SkyPilot, preserving Sinter's sampling, batching, and decoder integration while adding multi-node scale-out, spot-preemption resilience via S3-synced resume CSVs, and cryptographic-identity-based result merging. This archive contains the primary data underlying every measurement reported in the paper: stim_benchmark: 42-circuit rotated surface code threshold study on AWS m6i.4xlarge spot instances, single-node baselines plus 7-worker distributed runs in both "short" (10^7 shots / 10^3 errors) and "long" (5x10^8 shots / 10^4 errors) configurations, including a three-trial replication (R1, R2, R3) for the pre-built-image long variant. Table 1, Figure 1, and all wall-time claims in Section III derive from these CSVs. tsim_calibration: Single-node bloqade-tsim 5x5 (K, p) magic-state distillation sweep sampled on an AWS L4 GPU spot instance. K is the number of composed 5-qubit Reichardt distillation rounds; p is the physical depolarizing rate. tsim_grid_merged.csv is the canonical 25-cell result table that feeds the Section IV heatmap and the Section V per-cell wall-time figure. calibration_stats.csv records the first-at-K JIT compilation timings used to derive the per-K JIT-tax figures discussed in Section V. tsim_multicloud: Config B, the 5-worker cross-region AWS (L4/A10G mix) tsim distribution study at 10^7 shots per cell. Per-worker CSVs plus config_B_summary.md, which records the JIT fraction breakdown (48-61 percent per worker), the 6-attempt retry history for worker mcb_3, and total spot spend. The reproducer Docker image is published at Docker Hub: ohheyitsbrian/qec-kiln-worker-gpu:paper-v1 (digest sha256:d6bddf6eb2e84fa8c7fc247095d3e66a75397d6700db54ae75c2207cfc9bf6f7). Source code is at https://github.com/brianirish/qec-kiln (commit f91d565). Note on simulator version: the tsim measurements ran against bloqade-tsim 0.1.2. bloqade-tsim 0.1.3 ships a DEPOLARIZE2 calibration improvement that does not affect the paper's wall-time and JIT-tax claims, but may shift logical error rates for independent re-simulations.