As a policy of countercyclical demand management, Keynesianism is based on the idea that there are no automatic forces in industrial capitalism that reliably equilibrate aggregate supply and aggregate demand. In the absence of government stabilization policies, the economy will waver between inflationary periods of excess demand and depressed periods of inadequate demand. The main explanation for this instability is that private investment depends on long-term profitability expectations, but since aspects of the future relevant to profitability are fundamentally uncertain , these expectations are unanchored and conventional, inevitably subject to large collective shifts independent of current “fundamentals”.
Government spending can stabilize demand if G+I varies less over the business cycle than I alone does. For which it’s sufficient that government spending be large. It’s even better if G and I move in opposite directions, but the reason Minsky answered No to Can “It” Happen Again? was because of big government as such, not countercyclical fiscal policy.The focus on cyclical stabilization assumes that there is no systematic long-term divergence between aggregate supply and aggregate demand. But Keynes believed that there was a secular tendency toward stagnation in advanced capitalist economies, so that maintaining full employment meant not just using public expenditure to stabilize private investment demand, but to incrementally replace it.
Another way of looking at this is that the steady shift from small-scale to industrial production implies a growing weight of illiquid assets in the form of fixed capital.  There is not, however, any corresponding long-term increase in the demand of illiquid liabilities. If anything, the sociological patterns of capitalism point the other way, as industrial dynasties whose social existence was linked to particular enterprises have been steadily replaced by rentiers.  The whole line of financial innovations from the first joint-stock companies to the recent securitization boom have been attempts to bridge this gap. But this requires ever-deepening financialization, with all the social waste and instability that implies.
It’s the government’s ability to issue liabilities backed by the whole economic output that makes it uniquely able to satisfy the demands of wealth-holders for liquid assets. In the functional finance tradition going back to Lerner, modern states do not possess a budget constraint in the same way households or firms do. Public borrowing has nothing to do with “funding” spending, it's all about how much government debt the authorities want the banking system to hold. If the demand for safe, liquid assets rises secularly over time, so should government borrowing.
From this point of view, one important source of the recent financial crisis was the surpluses of the 1990s, and insufficient borrowing by the US government in general. By restricting the supply of Treasuries, this excessive fiscal restraint spurred the creation of private sector substitutes purporting to offer similar liquidity properties, in the form of various asset-backed securities. (Here is a respectable mainstream guy making essentially this argument. ) But these new financial assets remained at bottom claims on specific illiquid real assets and their liquidity remained vulnerable to shifts in (expectations of) the value of those assets.
The response to the crisis in 2008 then consists of the Fed retroactively correcting the undersupply of government liabilities by engaging in a wholesale swap of public for private liabilities, leaving banks (and liquidity-demanding wealth owners) holding government liabilities instead of private financial assets. The increase in public debt wasn’t an unfortunate side-effect of the solution to the financial crisis, it was the solution.
Along the same lines, I sometimes wonder how much the huge proportion of government debt on bank balance sheets -- 75 percent of assets in 1945 vs. 1.5 percent in 2005)contributed to the financial stability of the immediate postwar era. With that many safe assets sloshing around, it didn't take financial engineering or speculative bubbles to convince banks to hold claims on fixed capital and housing. But as the supply of government debt has dwindled the inducements to hold other assets have had to grow increasingly garish.
From which I conclude that ever-increasing government deficits may in fact be better Keynesianism – theoretically, historically and pragmatically – than countercyclical demand management.
 Davidson, Shackle, etc. would say nonergodic. This strand of Post Keynesian thinking often wanders beyond my comfort zone. This shift is ongoing, not just historical -- not only do capital-output ratios continue to rise in manufacturing, but we're seeing the "industrialization" of retail, health care, etc.
 Schumpeter is the only major economist to give sufficient attention to the sociology of the capitalist class, IMO. Marx's insistence that the capitalist is simply the human representative of capital is a powerful analytic tool for many purposes, but it leaves some important questions unasked.
 Here is another.