I=S for all transactions

But my most controversial charge was that Wren-Lewis (and by implication Krugman) forget that S=I.Money Illusion

I go with Scott. The idea is that I=S has to be the outcome in the model, if one assumes that I!=S, then one has a new model. Can't have it both ways.

Channel flow does have moments outside the model, sorry, I guess those periods radiate goods. Oct 2009? The model on that day was certainly different than the usual model.
The general question is, can one construct a single model always and everywhere consistent?  Not when I,S -> 0, because near zero the model blows,  and all models usually have a zero so the algebra has additive identity, and thus generating a robust language.

The only repair is a transform that moves the definition of I,S away from zero, and that's a new model.