Hilarious that every time, they responded in a way that was technically on-topic, but totally ignoring the actual questions being asked. Like someone found one word in OP's question, then mindlessly recited a random form response associated with that word.

this is, legitimately and without any exaggeration whatsoever, almost exactly how nearly everyone I speak to IRL as a homeless individual interacts with me. There are a few here and there that absolutely do make an attempt to have real discourse and actually discuss in context, but by an overwhelming majority, most of them just take the (to borrow from a response) SLM approach.

life is exhausting sometimes

I assume they’re not really paying attention to you, like if you were a child vying for attention. You’re not likely to run into many salient points speaking to someone who isn’t tracking the basic things an adult should do.

Not an uncommon experience on, say, HN, to draw an arbitrary example.

A Small Language Model...

If this happened today, I'll bet there is a LLM.

Reading their replies felt awfully similar to how LLMs get stuck in loops, repeating the same patterns over and over.

Chatbots from the time before LLMs.

To me, it didn't come across as deliberately evasive. It came across like tier 1 helpdesk not really understanding how a query fits into their pre-defined categories, and trying to be helpful anyway without really understanding the problem.

The later reply about it being because OCR does use what's below the line and it shouldn't be obscured looks like the ticket was escalated to someone who understood what was really being asked for.

Obligatory xkcd: https://xkcd.com/806/

Except that you might be missing the key fact: the letter is just a letter. It’s not a form that requests information from the recipient, and the recipient is not instructed to return it. Scanning it would be pointless, because all of the information on it was printed out by them!

Ooh! I wonder if most of the letters they send out are forms (expected to be returned), but all of their letters are on the same paper-stock, which is pre-printed with the (intended for forms) "don't write below" message.

Evidence in favor of this theory: the "don't write" message is in red text. It's cheaper to do two (or more) passes on one high-volume print run - and then single (B&W) impressions on the smaller runs for each individual letter or form - than it would be to do multiple passes for each and every order.

The support staff wouldn't be privy to these sorts of economic optimizations, so no wonder they couldn't give the guy a comprehensive answer.

That’s entirely plausible, and it was my guess too.