And the empirical examples:

Re; the "Blockchain wars" that resulted in the BCash fork in 2017. The majority of the hashrate had signaled support for Segwit2X. Economic nodes prepared a UASF (i.e., said "no fucking way") which resulting in most of those miners realizing that the economic nodes wouldn't recognize (i.e., "buy") their BTC that they mined in SegWit2X blocks so they tucked their tails and went along with with the Segwit soft fork instead.

Prior to that, March 2013 ... the netsplit that occurred as a result of a new release that unintentionally had a change to the protocol. Some miners had not been running the new v0.8 client, and majority of the hashrate was running v0.8. A netsplit happened as a result of that unintentional change, and the side with the miners running v0.8 had been solving blocks faster, and were six blocks (coincidentally) ahead. What to do? Well, that was decided after one question was answered: "What side is Mt. Gox on?" Mt. Gox was like 80% of the exchange market at the time. Mt. Gox essentially had the majority of the economic nodes (by quantity of bitcoin buy/sell activity), and after learning that Mt. Gox had not been using v0.8, the leading pool miner voluntarily abandoned the v0.8 side (including multiple blocks that they had mined), and redirected their hashrate to the same side of the net split that Mt. Gox was on (i.e., the pre-v0.8 side). Thus the miners did a "friendly" 51% attack to comply with what the economic nodes demanded.