Fowler's description is, as I see it, basically accurate. To be somewhat less gentle about it than Fowler is, microservices are pretty much exactly the original description of SOA -- the use of ESBs to integrate monolithic applications that he describes as often the focus of current SOA discussions was originally a style of initially integrating legacy applications in an organization that was doing enterprise-wide SOA conversion. Something like it (the connection may not be an ESB, per se) is pretty much an essential stepping stone in a conversion effort (and enterprises are notorious for taking the first enabling step and then not following through with that that step is intended to enable.)

If microservices become a popular focus of attention, don't be surprised to see the term drift the way SOA did for the same reasons, and see us back here in a couple decades with a new term and discussing how it (doesn't) differ from "microservices". Never underestimate the power of the enterprise to both dominate the literature and to sap the meaning out of terms.