instance-state in the node-state table (DynamoDB/SimpleDB) is a hardcoded literal, never reflects real instance status

mdb-update-state() in mlcmd/scripts/mdb-functions.xsh hardcodes instance-state to the literal string "running":

mdb-update state $1 state-instance $MARKLOGIC_INSTANCE instance-state running instance-time <[ fn:current-dateTime() ]>
Per the comment in ddb-functions.xsh line 315 ("Keep quiet this is run every minute from cron"), this field is refreshed periodically regardless of actual node health.

Observed impact: after scaling an ASG-managed cluster to 0 (full stop), the DynamoDB node-state item kept reporting instance-state=running with instance-time still advancing for several minutes past InService=0, then froze permanently at "running" with a stale instance ID once the instance was actually terminated. Anything consuming this table to infer real node liveness (automation, dashboards, other cluster tooling) will be misled -- there is no way to distinguish "genuinely running" from "was running as of the last cron tick, possibly hours or days ago."

Suggested fix: either rename the field to something that does not imply liveness (e.g. last-seen-state), or have it reflect an actual check (e.g. query the instance's real EC2/process status) rather than a hardcoded literal.

Start in mlcmd/scripts/mdb-functions.xsh at mdb-update-state(), then read the related comment around line 315 of ddb-functions.xsh and trace how instance-state is consumed. The issue needs an agreed replacement semantics—renaming the field or checking actual EC2/process status—before implementation can be considered done.