Match Qwen30 probe file descriptor limit

runs/frontier-fidelity-envelope-v1/run_qwen30_fixed_pd_pressure_probe.sh

@@ -31,6 +31,12 @@ SERVER_PID=""
 mkdir -p "${OUT}/provenance" "${OUT}/trials" "${FLASHINFER_WORKSPACE_BASE}"
 exec > >(tee -a "${OUT}/controller.log") 2>&1
 
+# Match the validated Qwen30 real-surface runner's file-descriptor budget for
+# vLLM's multiprocessing shared-memory transport. The failed r1 probe had the
+# default remote limit (1024) and stalled before KV-cache creation; r2 tests
+# whether this runner-contract difference is causal.
+ulimit -n 65536
+
 cleanup_server() {
   if [[ -n "${SERVER_PID}" ]] && kill -0 "${SERVER_PID}" 2>/dev/null; then
     kill -TERM -- "-${SERVER_PID}" 2>/dev/null || true
@@ -204,6 +210,7 @@ PY
   sha256sum "${BASH_SOURCE[0]}" "${CLIENT}" "${MODEL_ROOT}/config.json" > "${OUT}/provenance/input.sha256"
   "${VENV_ROOT}/bin/vllm" --version > "${OUT}/provenance/vllm.version"
   "${VENV_ROOT}/bin/python" -c 'import torch, transformers, vllm; print(f"torch={torch.__version__}"); print(f"transformers={transformers.__version__}"); print(f"vllm={vllm.__version__}")' > "${OUT}/provenance/runtime.versions"
+  ulimit -n > "${OUT}/provenance/open-file-limit"
   nvidia-smi --query-gpu=index,name,uuid,driver_version,memory.total --format=csv,noheader > "${OUT}/provenance/gpus.before.csv"
 
   declare -a ORDERS=("4 8 12 16" "16 12 8 4" "8 16 4 12")