Entirely relying on unit tests for development can never be the only solution, because tests are quite simply not the product, and they shouldn't be a substitute/workaround for fast iteration right on the product. Easier said than done of course, especially in such a complex environment as game development.
The future for fast compiled-code iteration in game development will most likely be hot code reloading, so that code changes are compiled and implanted right into the running game instead of requiring a full linker run and then getting back to the right place in the game to be tested.
I don't think anybody is ever suggesting this - but I do see people suggest the opposite: that all testing should be end-to-end "black box" testing, and unit tests are a waste of time. If you actually want to ship something that works reliably, you have to do both unit testing and end-to-end (integration) testing. I've never seen anybody sacrifice integration testing. I have seen them sacrifice unit testing.
The future for fast compiled-code iteration in game development will most likely be hot code reloading, so that code changes are compiled and implanted right into the running game instead of requiring a full linker run and then getting back to the right place in the game to be tested.