The migration review is the fourth one this quarter. Same room, same slide, same architect raising the same objection.
The vendor’s failover claims don’t hold under load. He said it in March. He said it in April. Nobody actioned it — the vendor’s own engineers had signed off, twice, in writing, and a signed attestation outranks an internal objection with no incident behind it yet.
In June, the failover doesn’t hold under load.
He is right. Everyone says so, generously, in the postmortem. He is also, by then, three months past the point where being right helped anyone.
Here is what being right, late, is worth.
Correctness is not the same as usefulness. A warning that arrives before the decision changes the decision. The same warning, arriving after, is not a warning. It is a receipt.
He was not ignored because he was wrong. The room heard him, weighed his warning against two written vendor sign-offs and a mixed track record, and made a defensible call that turned out wrong. A different architect raised a different objection three postmortems ago, just as confidently, and was wrong. The room doesn’t distinguish between that signal and this one. It just remembers noise.
That is the uncomfortable gap: a correct prediction is not, by itself, an actionable case for stopping a live migration. No amount of listening closes it. The fix isn’t better listening. It’s attaching a cheap, discriminating test to the objection before the expensive failure does it for you.
He was right in March. The organisation needed him right in a way it could act on. That is a different, harder thing to be.
— Signal / unfiltered observations from the field. Fortnightly.
#OrdoAnimi #EnterpriseArchitecture #ArchitectureLeadership #ITGovernance #ITStrategy #Execution #CIO #DigitalTransformation