And since when does a for-pay piece of software eclipse the original free version in popularity? If it did happen, then I'd assert that the "proprietary company" must have done a lot of work to make that possible. In that unlikely case, they must have invested a nontrivial amount in the creation of a powerful tool, and they deserve the profits it generates.
A far more likely scenario is that a proprietary company forks an MIT-licensed project, does something creative with it, sells it to a few people, and then the open-source community would copy the cool features and add them to the MIT project. The actual code to a new feature is typically not as hard to create as the design of the feature, which has already been done by our hypothetical proprietary company. Then the company will need to innovate again to stay relevant, and the free project can copy those features as well.
A healthy ecosystem can include some companies that don't share. It's all part of the network effect. If you can cite any situation where a community surrounding a free software library or other product has been killed or even harmed by a piece of proprietary software that is based on the same code, then you might have a point, but I don't think it's happened.
It DOES happen that proprietary software runs circles around the equivalent free software (Adobe Photoshop, for instance), but that's not because the for-pay software is stealing free code. It's because when you have hundreds of full-time developers you can produce a lot more than the same number of people can do in their spare time. And when those full-time paid developers can contribute to open source (as is frequently the case with MIT/BSD licensed projects), those open source projects flourish as well. Again, see LLVM/Clang.