Looking at the network topology of the segwit softfork, it looks horrible. The idea that there is a mixture of nodes, some with full network knowledge and some without full knowledge of it if they don't upgrade is asking for complications in my opinion. I wouldn't be surprised if some of franky1's concerns are true, but cannot confirm this as it is way beyond my technical knowledge and expertise in this area.
Perhaps this should have been implemented as a hard fork along with the 2MB block size increase as originally agreed. If the miners and exchanges are behind a hard fork, the old fork is a stuck difficulty chain with coins that cannot be traded.
Even if the future is that the main chain is more of a settlement chain, with lightning style networks providing retail functions, there still needs to be an ability to grow the chain to meet demand. Some alt coins (e.g. Komodo) are now using the bitcoin blockchain, increasing demand. I don't see why some algorithmic block size change can't be implemented. It could look at parameters such as high fee / low fee unconfirmed transactions, past block usage, orphan rates, etc. and slowly adjust, monitor effects and adjust. Then a sensible fee market could develop, rather than this pathetic fee race or drop out due to expense situation we have now due to a fixed block size. A fee market cannot develop on a full block chain, it just limits use when it becomes to expensive to transact.
I'm not surprised the miners are upset with core / blockstream developers for not keeping to their word and implementing segwit as a hard fork along with a much needed and conservative 2MB block size change. So perhaps it is stalemate until we see which economic power goes bust first.