fix: 用容器生命周期核实恢复并单列三份总体验收

审阅发现spawn后快照早于恢复实际执行，不能证明没有重启；改用杀前至目标结果落账的Docker历史事件，保留缓存完整性限制。逐份stats不再冒称任务书总体通过；独立验收要求三份不同研究、成功杀进程及不重叠区间。两项缺陷附回归，7项定向测试与check通过，独立复审无新问题；真实B尚待启动。

eval/forward/stats.ts

@@ -12,6 +12,12 @@ import {
   jsonFile,
   scoreBand,
 } from "../lib/forward.ts";
+import {
+  sameContainerThroughResult,
+  type ContainerEvent,
+  type ContainerSnapshot,
+  type LifecycleWindow,
+} from "../lib/forward-evidence.ts";
 import { finalSubmissions, ofKind, readLedger } from "../lib/ledger.ts";
 import { lightCheckUsage } from "../lib/light-checks.ts";
 import { referenceScores } from "../lib/reference-scores.ts";
@@ -117,16 +123,32 @@ const summaries = dirs.map((dir) => {
   if (existsSync(join(dir, "kill.json"))) {
     const evidence = JSON.parse(
       readFileSync(join(dir, "kill.json"), "utf8"),
-    ) as { run: string; candidate: string; beforeKillSeq: number };
-    type Container = {
-      Id: string;
-      State: { StartedAt: string; Running: boolean };
+    ) as {
+      run: string;
+      candidate: string;
+      beforeKillSeq: number;
+      container: string;
     };
-    const snapshots = ["before", "after-kill", "after-resume"].map(
+    const snapshots = ["before", "after-kill", "before-resume"].map(
       (part) =>
         JSON.parse(
           readFileSync(join(dir, `container-${part}.json`), "utf8"),
-        ) as Container,
+        ) as ContainerSnapshot,
+    );
+    const lifecycleWindow = JSON.parse(
+      readFileSync(join(dir, "container-events-window.json"), "utf8"),
+    ) as LifecycleWindow;
+    const lifecycleEvents = readFileSync(
+      join(dir, "container-events.jsonl"),
+      "utf8",
+    )
+      .split("\n")
+      .filter(Boolean)
+      .map((line) => JSON.parse(line) as ContainerEvent);
+    const lifecycleChecks = sameContainerThroughResult(
+      snapshots,
+      lifecycleEvents,
+      evidence.container,
     );
     const target = accounts.runs.find(
       (r) => r.plan.content.run === evidence.run,
@@ -144,16 +166,18 @@ const summaries = dirs.map((dir) => {
       evidence,
       killSeq,
       snapshots,
+      snapshotMeaning: "杀前、杀后、恢复启动前；没有把快照当成恢复完成的证据",
+      lifecycleWindow,
+      lifecycleChecks,
+      lifecycleEvents,
       charged: target.starts.length,
       resultSeq: target.results[0]?.seq ?? null,
       checks: {
         sameContainer:
-          snapshots[0]!.State.Running &&
-          snapshots.every(
-            (c) =>
-              c.Id === snapshots[0]!.Id &&
-              c.State.StartedAt === snapshots[0]!.State.StartedAt,
-          ),
+          Object.values(lifecycleChecks).every(Boolean) &&
+          lifecycleWindow.run === evidence.run &&
+          lifecycleWindow.resultSeq === target.results[0]?.seq &&
+          Date.parse(lifecycleWindow.since) < Date.parse(lifecycleWindow.until),
         chargedOnceNoRerun:
           target.starts.length === 1 && candidatePlans.length === 1,
         resultAndContinued:
@@ -211,6 +235,6 @@ jsonFile(join(output, "manual-review.json"), {
     "核对各目录launches.jsonl、machine.jsonl和父会话观察；上一份全部结束后才起下一份；不宣称快照排除任意外部进程",
 });
 console.log(
-  `统计：${output}；程序检查${summaries.every((s) => s.pass) ? "通过" : "未通过"}；质量判断仍待独立阅读。`,
+  `统计：${output}；所给 ${summaries.length} 份研究的逐研究检查${summaries.every((s) => s.pass) ? "通过" : "未通过"}，不是任务书16总体验收；总体程序验收另运行 node eval/forward/acceptance.ts ${output}；质量判断仍待独立阅读。`,
 );
 if (!summaries.every((s) => s.pass)) process.exitCode = 1;