Replication and MVCC

Well, their issue with "query cancel", which is what this particular problem is called, was partly that they were on 9.2, where hot_standby_feedback didn't work as well as it does now. For PostgreSQL today, there's an effective tradeoff, depending on whether you care more about avoiding query cancel or avoiding load on the master.

Note that the PostgreSQL community also agrees that having better logical replication (row-based) would also be a good thing, hence the "BDR" project: http://bdr-project.org/docs/stable/index.html

There are benefits to logical replication, but also drawbacks. For example, logical replication imposes a much higher query load on the replicas, making them less able to load-balance read workloads.