Sandbox upgrades aren’t always accurate tests. I know of more than one rollout that learned the hard way they had stuff hardcoded to reference servers in the sandbox.
Right, so you roll out your upgrade in the morning . . . on a Wednesday.
You don’t roll your upgrades out on say, Friday after close of trade and then run your regression test set against the upgrade, or a simulated trading day against the upgrade.
Why would you do that - every other business in the US with any sort of IT competence would do that.
That’s why their explanation is BS. The NYSE has some of the best IT folks in the business. They didn’t roll an upgrade out in the AM on a trading day. They wouldn’t do that.
They would sandbox it in the middle of the week, run the regression sets and the trading simulator at it, and then you deploy it Friday at 10PM after the after hour trades are in.
Smells like caca, their explanation. Especially with all their redundancy and fail over.