Sure, there might be currently contraints, but I think it is quite possible training will get optimized over time or crowsourced training can be organized.
Without fully end-to-end open source models you are still at the mercy of the model provider to keep providing updates, you have no idea what garbage they trained the model on & can't fix that, not to mention might end up getting sued for using the open weigth model once all those "AI stole my data" lawsuits are finally decided.
You will not have that control. Even if everything were open source. This is because you don't have 100 million dollars of compute.
There, the difference for almost everyone is negligible.
> Without fully end-to-end open source models you are still at the mercy of the model provider to keep providing updates
No. Because you can post train it. And even if you could train the whole thing again, but with slight changes, once again, you aren't spending the hundred mil in compute to change it only a little bit.
Instead, you'll do post training like everyone else does.