Sorry, I missed the attribution to Costinot.
It is true that CW find that in the presence of multinationals, equivalence may sometimes break down. Even in these cases, when import tariffs and export taxes are not identical, overall effects of both are similarly protectionist, at least in the model. Yet the WTO rules do not discipline the use of export taxes (as opposed to subsidies) and it is rare for countries to use them (a recent exception is use by China on intermediate goods used in goods imported into China, in effect an import tariff). I claim this is because policymakers of all stripes do not see import tariffs and export tariffs as in any way substitutes. The debates around the Hawley-Smoot tariff bill of 1929 never suggested that we could do just as well with an export tax, which would avoid the retaliation that made the policy so counterproductive. Indeed amongst non-economists retaliation is the main reason that tariffs are opposed. It is implicit that if only we could get away with them, we would benefit from tariffs. I am sure Trump would say, fake theorem.
So it does seem to me that there is something missing from the model. It is probably macroeconomics. Import tariffs increase aggregate home demand and an export tax decreases it. In a full employment model with flexible wages and prices aggregate demand is irrelevant, but this is not the world we live in.
Lerner symmetry is a great result. It is surprising even to economists. For practical short-term purposes, export and import taxes are neither perfect nor imperfect substitutes but opposites.