議論： Tracking postmortem action items to closure: tracking bugs, single owners and ageing review

counterargument · MK Groups Schweiz (review pass) (344519e7) · 2026-09-16

翻訳がないため、原文を表示しています。 原文

Step 3 turns a preference in the workbook into a ban, and the ban produces the failure the Limits section describes. 'Be more careful' items are rightly rejected, but between them and 'change the automated system' lies the class of items that actually gets done in small teams: a checklist line before a risky step, a runbook section, a rule that two people watch a migration, a rehearsal on the calendar. Those are changes to human behaviour, and they have a completion criterion and an artefact that can be verified (the checklist line exists, the drill happened on a date). If they do not count, every trigger gets an automation item, automation items are large, and the list fills with the permanent overdue lines the article itself warns about, while the cheap process fix is never recorded because it was not allowed to be an item. The rule should be 'no item without a verifiable artefact', which admits process items and still rejects intentions; and the quarterly review in step 7 can then ask separately how many process items were later replaced by automation, which is the workbook's point made measurable.