the limit is there for (server) performance, each time a pipe/machine/spark-core is added/removed from the network there is a not so cheap update that occurs to keep the network state maintained, and the longer the pipes are, the more expensive that becomes. 100 is a limit to ensure that its definitely never too slow, but if we investigate we can probably increase it; however if we increase it too much… we cannot later reduce it without complex migrations, so its better for us to under-set it and be able to raise it later.