My questions for DeLong sparked two good non-DeLong questions in the comments:

From Steve Roth:

Employers *are* more likely (than employees) to put the money under their mattress.

Perhaps not literally. But actually/practically.

Employers spend a smaller percentage of marginal earnings on immediate consumption (which would increase short-term AD).

They may spend a smaller percentage on immediate consumption, but AD isn’t just consumption spending.  It’s also investment spending – and remember that due to credit market imperfections, cash flow is a good predictor of business spending.  See e.g. here.  

Furthermore, my “mattress point” is that the direct beneficiary of the wage cut doesn’t need to personally spend it for AD to rise.  He puts it in the bank, and the bank lends it to someone who will spend it.  (It’s true, of course, that banks like to build up excess reserves during recessions, but this doesn’t mean that they don’t re-lend the marginal dollar deposited).

From david [not Henderson]:

I thought Keynesians (new, neo, paleo, etc.) believed that, due to
downward price and wage stickiness, we get quantity adjustments in a
recession, i.e., falling output and rising unemployment. But, once
we’re in the recession, the argument seems to be that the quantity
adjustments would be even bigger if prices/wages weren’t sticky(!).

Nope.  In standard New Keynesian models, if a magic wand makes prices and wages flexible, idle resources go right back to work.  See e.g. Mankiw’s menu cost model.  The tradition of blaming recessions on price and wage rigidities, then warning against the evils of price and wage flexibility, is firmly rooted in what I call “dinosaur Keynesianism.”  But I can’t remember seeing any modern macro model that works that way.