Coming from the ops side of the fence, who should be opposed to this, it's also often the only way things get done.

I've worked a couple places that had an unspoken Catch-22 mantra: "ye shall not invest resources in apps that are not business critical".

So the only way to build something new was to skunkworks it until something major depended on it and you could get resources to actually productionize it.

A smart ops team should have a semi-standard "skunkworks to production" pipeline.