And yet... from people who never pay the bills for anything, all I ever hear is 'space is cheap, computing is cheap, blah blah blah'. It's all 'cheap' until you have to pay for it.
If you are getting anything out of it net greater than it and there isn't a better use for it keeping it is worth it.
TYPICALLY the ones I hear that from are developers who aren't in charge of money/budget, and usually don't ever think about the cost of anything. The 'shiny/new' syndrome, coupled with little oversight, leads to nasty $ surprises later on.
And yep, "programmer time is more expensive" all night long. But at some point, other expenses are coming out of someone's budget someplace. Just because it's less than some other line item doesn't mean those costs can't add up to more than was expected (or needed, or budgeted for).
It isn't free, but it's certainly cheaper than killing your entire company.
Being cheap about this is just likely to lead to more bugs and cost in debugging later anyway. There are tradeoffs and exceptions, but there's a reason the saying exists.
This is why businesses that deal with information need staff who are able to reason about the lifecycle of information.
Not making a reference count and respecting it killed the company using one store of duplicate data, if anything, prolonged the company.
But I would guess bandwidth dwarfs storage costs here, so it does seem odd.
Slow-to-retrieve backup services like S3 Glacier are usually much cheaper, too.