If doing so saved them $50 on their BOM for the APU/RAM (net of increased cost of cooling solution), it's not a bizarre decision at all. If Sony goes on to sell 100m PS5s this generation, that $50 savings will yield an additional $5B in income (not revenue) over the 6-7 year life of the console. Those are huge numbers.
Additionally, those BOM savings can allow you to sell your console at an accordingly lower retail price. "Well instead they should just make it stronger and take an additional loss. They'll make it up on software!" Maybe, but there's a limit to that line of reasoning, and if taking a $100 loss on a $500 machine would allow them to sell at $400, wouldn't taking a $100 loss on a $450 machine allow them to sell at $350? (Obviously there's a limit as to this line of reasoning, since one might argue that they should just slap a new coat of paint on a PS4 Pro and sell it for $200 and call it a day. But I don't think 9.2TF is way out of line from a next-gen performance perspective - especially given many/most expected both consoles to be targeting 9-10TF this gen.) Additionally, Sony has long been focusing on the benefits of next-gen that aren't focused on raw graphics power - SSD, hw-assisted RT, better CPU, controller haptics, etc. - that it makes sense that they might feel like they didn't have to really stretch themselves on the TF front.
I'm personally 50/50 on the whole thing, but I don't think they'd be making an absurd decision either way.