The biggest mystery to me is how we arrived at this situation, where Gavin and Mike (felt they) had to resort to the fork solution mess...
Sure, a certain number of maxblockheads resisting any and all changes to the protocol I can see as being unavoidable. But why Gavin didn't manage to convince a majority of the core devs to support some adjustment is still not entirely clear to me.
Either Gavin tried to push exclusively his idea of an adjustment (which, admittedly, can seem a bit drastic) - in that case, it's a failure on his side to compromise. Alternatively, there was a consensus by the rest of the team that the current limit better not be touched at all right now (in which case, Gavin did the right thing, even if the right thing is a huge mess as well).
Seeing that Blockstream rejected even Jeff's 2 MB attempted compromise proposal, it is undeniable that they don't want ANY incerase in the block size limit at all. Their concerns about the effect of big blocks on nodes are just excuses. In fact, they have clearly stated their plan for bitcoin: get the network into congestion so that users are forced to compete for space in the blocks by raising the fees. Until most peer-to-peer traffic is driven out to off-chain solutions, leaving only high-value traffic on the blockchain -- such as settlements between hubs of some "overlay network". Some Blocstream devs mumbled about fees in the range 10--100 USD per transaction...
Blockstream's plan makes no sense to me, and neither to Gavin and Mike. I thought that Blocksream may have some secret agenda that justified wreching bitcoin, but maybe they are just unable to do such quantitative analyses and genuinely believe that their plan will work. Gavin says that he has tried to convince the Blockstream devs to increase the limit for a long time, byt they refused.
The 20 MB of Gavin's first proposal, and the automatic increases of BIP101, were clearly motivated by the FUD that Blocksrteam planted about hard forks (which in fact
are not technically different from the soft forks that Blockstream likes). A single increase to 8 MB, that the Chinese pools considered acceptable, would have been good enough; but critics would say that it would require another hard fork, a few years later.