Housing's "more roads" is simply
density.
In transport, more roads means more capacity to carry traffic, and hence, more induced traffic. The death-spiral effect comes from the fact that each additional car on the road produces far more congestion than its operator alone experiences (negative externalities). True costs aren't born by the operator, so gridlock ensues.
For housing, or rather, urban development generally, the challenge is that denser cities with more inhabitants (generally) give rise to greater opportunity. This is where the action is, for whatever definition of action you have.
I'm not entirely sure what produces the death-spiral of cities pricing themselves out of their own housing markets, as SF is doing, but I'm inclined to think that the general trend of holders of investment assets to seek inflation of those assets' values is a key element. That is, you have a house, you've sunk a lot of money into it, you want its value to increase (hey! free money, right?). This turns into collaterisable assets for loans or other opportunities as well.
There's also the fact that high-value urban real-estate markets make for good options to bank wealth appropriated elsewhere, or for money-laundering. Values are high, tend to remain high, the market is liquid, and urban cores are far more stable for real estate prices than outlying areas. Hence, the housing stock value is bid up further.
Land value tax might help address that. Something I'm looking at. Originally suggested by Adam Smith and David Ricardo, though championed by Henry George.