Release notes 0.1.164
28 October 2026: A certified runbook demoted for a fix that did not hold, before the outside write that explains the recurrence had reached the change ledger, gets its certification back when that…
Release notes · Updated 2026-10-28 · Describes release 0.1.164
0.1.164 — 28 October 2026
A certified runbook demoted for a fix that did not hold, before the outside write that explains the recurrence had reached the change ledger, gets its certification back when that write arrives, and the recurrence's change request that nobody has requested yet is classed again. The Changes list loads faster, an identifier in a table cell breaks at its separators first, and the list of EC2 instances not managed by SSM fits its card. No schema change, no new setting, route, permission or chart value; two new audit actions (lkg_record.outside_explained, change.class_restored) and a restored field on route.deviation_explained. On the routes list, a runbook demoted after the upgrade keeps the whole certification it lost under certification.held, and its demoted marker names the recurrence (by_incident) and any later deviation (again). Nothing to do for the upgrade (see the upgrade row).
A late outside write and the runbook it demoted
- The runbook gets its certification back. Since 0.1.158 a fix that did not hold demotes the certified runbook that made it (Demote a certified runbook when its fix deviates,
aiops_demote_on_deviation, on by default) unless the change ledger holds an outside write on the same resource between the fix and the recurrence. The mark and the demotion are written when the recurrence opens, while a cloud audit log delivers a write minutes later (CloudTrail 5 to 15 minutes): on 27 October an operator'sStopInstancesreached the ledger five seconds after the mark, and the runbook stayed demoted. After every audit harvest that records a cloud write, the platform now reads again the did-not-hold marks of the last 24 hours that name no outside write, newest first, up to 20 still-open marks a run. A mark the new write explains gets the audit row Recorded fix recurrence explained by an outside change and counts as a mark that named its write at once: the record reads recurred after an outside change, and the mark no longer counts towards Automatic fix blocked or the not-good state. The runbook that this recurrence demoted is certified again with the certification it held — certifier, revision, self-certification, note and permissions — audited Runbook certification restored — an outside change explains the recurrence (route.deviation_explainedwithrestored: true, on the recurrence incident's Timeline), and it leaves Needs you. The mark itself stays: the condition did come back. - The recurrence's change request goes back to standard. A change request the recurrence's analysis drafted while the runbook was demoted was classed
normal. When the runbook is certified again this way, each such request that nobody has requested yet (no approval asked or given) is classed again by the drafting rules; when the certified runbook covers its fix step for step it becomesstandardat that class's tier, audited Change class restored — the runbook was certified again (change.class_restored, with the class and tier before and after). A requested, approved, released or emergency request is never touched. - When the certification stays away. The restore gives back only the demotion that this recurrence's own mark caused. It is refused, and an approver certifies the runbook again as before, when the demotion came from an earlier recurrence nothing explains; when the runbook was hit again while demoted (a failed validation, another recurrence nothing explains, a rollback); when a person disabled, edited, certified or withdrew it, or chose Keep it manual, in the meantime; and for a demotion written before 0.1.164. A problem whose automation candidate the recurrence dropped keeps it dropped. The route row and the runbook catalog still word a restored certification as a kept one (… certification kept); the audit row and the incident Timeline say restored.
Pages
- The Changes list loads faster. The list masks each row's credentials in every field it shows; it now reads the row's credentials once per row instead of once per field. On 0.1.163 a live workspace's remediation list took about 9.6 s, nearly all of it this masking; on a 25,000-row test ledger a 50-row page of it went from about 2.9 s to 0.45 s. What is masked is unchanged.
- Identifiers break at their separators first. On the live 0.1.163 read-back a CloudFormation name split mid-part in the Knowledge base list, a sha256 digest inside its hex in a planned change's scope table, a region mid-word in a change's resource list (sou|theast), a role ARN in the Dashboard's service cell (r|ole), and an EC2 XML reply widened the Audit log's detail column. These cells, and an audit row's raw record, now break an id, ARN or region after a dot, colon, slash, hyphen, underscore, bar or comma first, as list tables do since 0.1.163; only a cell holding a run of 24 characters with no separator may also break inside that run. A sha256 digest in an Artifacts title or a scope resource shows its first 12 hex digits and …, with the whole value on hover.
- The unmanaged-instances list fits its card. Under Security posture → Audit & coverage, Scores, the Patch compliance card's EC2 instances not managed by SSM list, opened, ran 20 px past its card at 1280 px and scrolled the whole page sideways at phone width. An instance's Name is now cut to its column with the whole name on hover, its id, account and region break at separators, and below the widest layout the table scrolls in its own region, never the page.