This is a very good summary of the situation, thank you for posting this. However, one critical issue is omitted in the above: The suggested approach (SegWit),

This is a very good summary of the situation, thank you for posting this. However, one critical issue is omitted in the above: The suggested approach (SegWit), albeit probably more by coincidence than by intention, also corrects a shortcoming of the existing transaction protocol that lets you reuse parts of the computation when trying to find a suitable block, leading to a net gain of ~20% hash power over other miners if you exploit it. Critically, said approach ("AsicBoost") is patented! Therefore, not all miners can use the approach, leading to a centralization of hash power in contrast to the original intention of Bitcoin. Miners who currently benefit from this oversight in the protocol have no short-term interest in removing it via a change of protocol headers that SegWit entails, in addition to the other reason you mention, i.e., higher transaction fees if there is a backlog due to small blacks.