Segment 32
In this sub-session, we transitioned from the intense engineering of Phase 12's memory backpressure fix into consolidation and performance characterization. We formally recorded the Phase 12 split GPU proving API, use-after-free fix, early a/b/c free, channel capacity auto-scaling, and permit-through-send fix into the project's main documentation (`cuzk-project.md`), updating the summary timeline, stopping points, E2E test results, and file references. We also revised `cuzk.example.toml` with optimal defaults (`gpu_threads=32`, `partition_workers=12`). Then we conducted a systematic low-memory benchmark sweep across nine configurations (`pw=1/2/5/7/10/12` × `gw=1/2`) to characterize the engine's memory footprint and throughput for smaller systems. The results revealed a clean linear memory scaling formula (`~69 GiB baseline + pw × ~20 GiB`) and showed that `gw=2` provides no throughput benefit below `pw=10` due to synthesis starvation. Concrete deployment guidance was established for 128 GiB through 768 GiB systems. All changes were committed as `9bb657e5` on the `feat/cuzk` branch.