Issue

The metric arangodb_aql_current_query returns a huge value close to uint64 max, while there are no active queries (queries.current() returns empty).

This appears to be a misreporting issue, possibly due to counter overflow or internal state corruption.

Example output:

arangodb_aql_current_query{role=“SINGLE”} 18446744073708371286

Environment:

• ArangoDB version: 3.12.4 (Community)
• Architecture: SINGLE
• OS: Ubuntu 22.04 (if applicable)
• Cluster mode: No, single instance
• How started: systemd service (not in Docker)

Additional info:

• queries.current() returns empty array
• No recent errors or warnings in logs (journalctl -u arangodb3.service)
• Metric /admin/metrics is used directly by Prometheus

Expected Behavior

The metric should report the actual number of running AQL queries (typically 0), or at most a reasonable number, not 18446744073708....

💬 Let me know if further debug output or /metrics dump is needed.