At my last job, the (rather dedicated) developers would often push code to production at 1am, give or take an hour.

As support, I would ask them "Please don't do that. You don't watch to make sure it works okay"

They would respond 'It's okay, because we'll be back online at 8am'.

"Yes, but our primary customers are farmers, who check our SaaS at 6am to plan their day. It's the point of our software."

'It's okay, because we'll be back online at 8am'

"I don't think you understand how this works..."

It's an interesting problem to optimise for.

When is the best time to push new changes to production based on probability they will fall over, usage patterns and day of the week.

I'm relatively lucky as the stuff I do is mostly for businesses so the (obvious to me) push point is 8am Saturday morning, that gives me two days to sort out any problems and there is always some low level usage.

I think vacri hit it with the knowing your audience and what times they are active. If you run a true 24-hour global service then you need to think hard, but a lot of systems have a very select audience that can be properly scheduled. I would imagine midwest farmers would accept updates done Sunday mornings during church hours maybe slipping later into the afternoon during sports.

Sounds like a support organization that should have had the developers' home phone numbers.

Except in those cases we would hope they implement the feature in question and in this case we can only hope he has the good sense to leave it off.