After upgrading to OpenClaw 2026.6.1 (build 2e08f0f), per-run cron history is no longer persisted.

On the affected host:

• ~/.openclaw/cron/runs/history.db and ~/.openclaw/cron/runs/index.db exist but are 0 bytes (ctime = the upgrade date; never written since).
• All legacy ~/.openclaw/cron/runs/<uuid>.jsonl files were renamed to <uuid>.jsonl.migrated (100 present); there are 0 live .jsonl files.
• openclaw cron runs --id <id> therefore returns no historical rows.
• openclaw cron list --json job state (lastDurationMs, lastStatus, consecutiveErrors) is still updated, so scheduler state is intact — only the per-run history store is empty.

Expected: new cron runs append to history.db / index.db after migration.
Actual: the SQLite stores are created empty and never populated; run history is lost going forward.

Environment: OpenClaw 2026.6.1 (2e08f0f), Linux (Debian/Ubuntu), node v22.22.1, gateway under systemd (user unit), ~80 cron jobs.

Workaround in place: we mirror run outcomes into our own logging DB via a cron wrapper and read openclaw cron list --json state for health.

Possibly related (same release, lower priority — mentioned only in case it helps):

• Diagnostics "critical" heap threshold appears calibrated to a ~3GB heap while running an 8GB heap (NODE_OPTIONS=--max-old-space-size=8192), producing ~280 false-"critical" diagnostics over 3 days with NRestarts=0.
• Per-job failureAlert{} exposes only a per-job cooldownMs — there is no global cross-job storm-cap — so a single upstream provider outage floods the alert channel from many jobs at once.