In my own experiences I've found that to be mostly a myth. Most researchers are too focused on grants, running labs, etc to actually get proper collaboration and support done. It seems the "good researchers" who are actually worth paying have built a career from advancing their one topic from day one, lone wolf style.
Probably 50% of that non-groundbreaking work you mention is completely worthless, and the field grew in spite of those wasted resources, as dnautics says.
Even then, if the system was honest with individuals about their science abilities (not everyone can be a superstar), then there would be much benefit in preventing these people from advancing on their own in a broken system, and giving them the opportunity to have an actual job in the groundbreaking areas as a tech or such.