I've never really seen tech debt cause an immediate, hugely-costly single event where you could say "see, that was the risk we were taking."

I've seen it frequently slow feature development down, though.

But... I've also seen a lot of rewrites fail to improve feature development speed.

So until we, as a discipline, can quantify and predict development speed w.r.t. shitty vs good code, it's going to be a tough conversation that'll rely on persuasion and gut estimates.

We normally can't even predict how expensive (time-consuming) doing that rewrite that we want to do would be! Let alone the benefit!