Persistent audit log around cluster operations

[benbardin]

In reviewing a recent customer incident, we realized we were dependent on the customer's logging of key operational events like node startup, shutdown, and upgrade. Future investigations would be aided by a CRDB-owned persistent log of these operations across the cluster. Ideally, this log would store:

• Node ID
• Timestamp
• Event type (startup/cluster join/drain/shutdown)
• SHA/version

Ideally, this log would never wrap - or if needed, would wrap on a Very Long timescale.
Also ideally, this log would be kept on every node for every node, so the record could include decommissioned nodes.

(Full context here.) Thank you!

cc @arulajmani @ajstorm @nicktrav

Epic CRDB-42978

[benbardin]

Oh, I see much of this information is in system.eventlog.txt! That's great. So the request here would be to expand that log slightly, and verify its wrapping behavior.

[benbardin]

More details in https://cockroachlabs.slack.com/archives/C07VDN3CA3U/p1732294864390839