fix: recognize clean one-shot completions in health gate and drift (#1691)

* fix: recognize clean one-shot completions in health gate and drift

Treat exit 0 with restart policy no/absent as successful completion so
init and migration jobs no longer fail post-update observation or show as
service-missing, while long-running restart policies still fail closed.

* fix: ignore residual health on clean one-shots and honor deploy.restart_policy

Completed exit-0 jobs with no-restart intent no longer fail the health gate on leftover starting/unhealthy state, and Drift treats deploy.restart_policy with Compose precedence so any/on-failure services are not mistaken for one-shots.

* fix: require explicit Compose restart no for one-shot recognition

Docker inspect reports restart no for both intentional jobs and bare services that omit restart, so Health Gate and Drift now require declared restart:""no"" (or deploy.restart_policy condition none) and load Compose intent once per gate.

backend/src/utils/oneShotCompletion.ts

@@ -0,0 +1,77 @@
+/**
+ * Helpers for clean one-shot completion and Compose restart-intent normalization.
+ * Used by Health Gate and Drift service-presence only; does not change bulk
+ * status, Auto-Heal, or atomic-deploy helpers.
+ */
+
+/**
+ * True only for an explicit Compose `restart: "no"` (after normalization).
+ * Absent / null / empty do not qualify: Docker reports HostConfig restart
+ * "no" for both intentional one-shots and bare long-running services that
+ * omit `restart:`, so consumers must pass declared Compose intent, not inspect.
+ */
+export function isNoRestartPolicy(policy: string | null | undefined): boolean {
+  return policy === 'no';
+}
+
+/**
+ * Normalize Compose restart intent for Drift / declared-model consumers.
+ *
+ * Compose precedence: when `deploy.restart_policy` is set, its `condition`
+ * wins; otherwise the service-level `restart` field is used. Conditions map to
+ * Docker-like policy names so {@link isNoRestartPolicy} stays the single gate:
+ * `none` → `no`, `any` → `always`, `on-failure` → `on-failure`. Missing
+ * condition defaults to `any` (Compose default). Unknown shapes fail closed.
+ * Absent service restart (no deploy policy) stays null and is not one-shot eligible.
+ */
+export function normalizeComposeRestartIntent(
+  serviceRestart: string | null | undefined,
+  deploy?: Record<string, unknown> | null,
+): string | null {
+  if (deploy && Object.prototype.hasOwnProperty.call(deploy, 'restart_policy')) {
+    const policy = deploy['restart_policy'];
+    if (policy === null || typeof policy !== 'object' || Array.isArray(policy)) {
+      return 'always';
+    }
+    const condition = (policy as Record<string, unknown>).condition;
+    // Missing, empty, or non-string → Compose default `any` → always.
+    if (typeof condition !== 'string' || condition === '') {
+      return 'always';
+    }
+    switch (condition.toLowerCase()) {
+      case 'none':
+        return 'no';
+      case 'on-failure':
+        return 'on-failure';
+      case 'any':
+        return 'always';
+      default:
+        return 'always';
+    }
+  }
+  if (serviceRestart == null || serviceRestart === '') return null;
+  return serviceRestart;
+}
+
+export interface OneShotCompletionInput {
+  state: string;
+  /** Exact Docker exit code; null means unknown (fail closed). */
+  exitCode: number | null;
+  /**
+   * Declared Compose restart intent after normalization (`"no"` for explicit
+   * one-shots / `deploy.restart_policy.condition: none`). Do not pass Docker
+   * inspect HostConfig values here.
+   */
+  restartPolicy: string | null | undefined;
+}
+
+/**
+ * True only for an exited container with exit code exactly 0 and explicit
+ * declared restart `"no"`. Null/absent restart, null exit codes, and restarting
+ * policies never qualify.
+ */
+export function isCleanOneShotCompletion(input: OneShotCompletionInput): boolean {
+  return input.state === 'exited'
+    && input.exitCode === 0
+    && isNoRestartPolicy(input.restartPolicy);
+}