Any argument that the Kenyan government could not accomplish most of the same good by printing pesos will mean that the primary mechanism of GiveWell's effectiveness must be the U.S. dollars being exchanged for the pesos on the foreign currency market.
[...]
To reiterate, the sum of the good accomplished by GiveDirectly should equal the good accomplished by the Kenyan government printing pesos and distributing them to the same recipients, plus the good accomplished by GiveDirectly purchasing pesos on the foreign exchange market using US dollars and then burning them.

One thing that's missing, when you decompose GiveDirectly into money-printing and distribution by the Kenyan government plus foreign exchange purchases by Americans, is the expectations channel. If the Kenyan government actually printed a bunch of money and handed it out to poor households, that would signal the Kenyan government's willingness to fund projects by printing money which would raise expectations of future inflation and increase macroeconomic uncertainty. (In special situations helicopter drops can work as monetary policy, if done by a central bank for a well-defined monetary purpose, but helicopter drops as business-as-usual anti-poverty policy by the central government are a bad sign.) It would be a pretty odd situation if foreign donors were working to maintain Kenyan macroeconomic stability while the Kenyan government was printing money to give away goodies.

A problem with using foreign exchange purchases of a developing world currency as an anti-poverty intervention is that the financial benefit goes to holders of that currency in proportion to the amount that they hold. So I suspect that the effect of giving the money to selected recipients is large, since it turns the intervention from one that especially benefits richer Kenyans into one that especially benefits poorer Kenyans.