I am not sure what you seek - a faster response from RE for pings? This is what you are testing as any local IP associated with the switch, is basically the RE/CPU. All that has happened in new products and newer Junos is that priority of processig for ICMP has been reduced versus other processes. So when there is a delay, it is because the RE/CPU is busy doing something else more important.
This test is NOT a test of the switches ability to pass real transit traffic and internal switch latency. This type of test goes back to mid-90s when switching devices were purely CPU based and therefore this type of test could show product latency and performance.
If you want to actually test for switch latency, you must ping or pass traffic through the switch, NOT to the RE. The RE/CPU is not involved in any transit packet processing. Packet could be copied to RE/CPU but the packet still passes through the switch with same performance and latency as non-copied to RE/CPU packets.
This has been the design of switches and RE/CPU/Sups/etc. for a long, long time.
BTW, EX4600 and QFX5100 are basically the same switch design, with same CPU, etc. so they should operate similarly.