Environment: Azure Container Apps Job (D32-benchmark, 32 vCPU / 64 GiB), westus3
ClickHouse Cloud Production tier (3 replicas × 16 vCPU × 64 GiB = 48 vCPU / 192 GiB)
Connection: HTTPS, Compression=true, set_async_insert=1, set_wait_for_async_insert=0
PBP sink: clickhouse (direct bulk-copy, 4-way parallel per chunk)
Merge parallelism cap: 16 (was 4 on smaller tier)
Job execution: oc-exp-1k-p32-xwqft3i
Image: commit ba0bb337 (does NOT include season accumulator — that landed in 363a7726 after this job started)

Backend: clickhouse, Scale: 100000 worlds, Environment: local
Streaming mode: totalWorlds=100000, chunkSize=500 [parallel=32]
PBP sink: clickhouse at ./pbp-parquet

Streaming Run Report
--------------------
Total wall time:         9886.65 s
Simulation time:         1462.17 s
OC write time:            338.90 s
PBP write time:          7955.54 s
Merge time:               122.02 s
Peak working set:        46827.1 MB
Chunks completed:            200
OC rows (merged):        244,800
PBP rows (written):   4,198,546,157

Notes:
- FIRST SUCCESSFUL 100K full-pipeline run against ClickHouse Cloud.
  Previous attempt on Production Scale (2×4) tier hit the 4h replica
  deadline mid-PBP-write (super-linear scaling past 10K).
- 3×16 tier resolves the PBP scaling problem: throughput of 528K rows/sec
  at 100K vs 513K rows/sec at 10K — essentially flat.
- PBP write is now the dominant phase at 80% of wall time (2h 13m of the
  2h 44m total). Sim + merge + OC write combined are only 32 min.
- Peak client RAM 46.8 GB / 64 GiB — comfortable margin.
- PBP row count 4,198,546,157 matches the 10K×10 expected value.
- Cost estimate: ACI 2h 45m × ~$1/hr = $2.75; CH Cloud 3×16 × ~$8/hr =
  $22; total ~$25 per 100K full-pipeline run.
- DOES NOT INCLUDE SEASON CONTEXT — wiring for that landed in commit
  363a7726 after this run started. Next validation: 10K full-pipeline
  including season context, then 100K.
