ServiceRadar v1.4.92

This release closes abandoned sweep executions and speeds up mapper
promotion while ingesting sweep results.

Fixes

• A maintenance job runs every five minutes to fail sweep executions
  abandoned by their agent, so they no longer remain in Running Scans
  indefinitely. A disabled group's execution is eligible after ten
  minutes without progress. Other executions become eligible after
  three times the group's interval, bounded between thirty minutes and
  six hours; cron-scheduled groups use six hours. Each pass handles up
  to 200 executions, oldest heartbeat first, and leaves an execution
  alone if a result chunk advances its heartbeat during the check
  (PR #5604).

Performance

• Sweep result ingestion writes mapper promotion state in batches of
  up to 1,000 devices per SQL statement instead of updating each device
  in a separate transaction. The database merges metadata without
  overwriting unrelated keys, touches only live devices, and locks rows
  in the same order as inventory sync writes. Cache invalidation,
  device update notifications, cooldowns, and dispatch payloads are
  preserved (PR #5605).