I think many people on HN are used to throughput oriented horizontally scalable services, where GC is not particularly relevant because: a) You don't care abou

I think many people on HN are used to throughput oriented horizontally scalable services, where GC is not particularly relevant because: a) You don't care about pause times since latency isn't your key metric b) You can almost always just throw more money at compute power to increase throughput It's almost always the case in such a system that your bottleneck is on the databases that you're pushing data through, as your main approach of "scale horizontally" typically leads to those sorts of bottlenecks.