Very interesting test, but not very conclusive. From video history of the same test performed on different devices by C4ETech:
1:05 - 1:54 - 1st and 2nd round - Galaxy S7 (exynos, 4Gb)
1:01 - 1:35 - 1st and 2nd round - Galaxy S7 Edge (exynos, 4Gb)
1:24 - 2:00 - 1st and 2nd round - HTC 10 (SD820, 4Gb)
1:05 - 2:10 - 1st and 2nd round - LG G5 (SD820, 4Gb)
1:10 - 2:22 - 1st and 2nd round - Oneplus 3 (SD820, 6Gb)
If you compare those numbers, you will see that:
1) +/- 5 seconds don't really mean anything. S7E vs S7 with same CPU and same ram in 2 tests gave different results in the 1st round.
2) Memory management can be luck of the draw and depends on other factors. Again difference is S7 vs S7E clearly shows that one app is still in memory in one test and not in memory on the other, even though hardware and software are almost the same (if you don't consider S7 and S7E to be all that different). Otherwise you might as well call S7E "awesome" phone and S7 "crappy" phone.
3) Speed of all those (except HTC10) is almost the same judging by the 1st round. HTC10 is just a bit slower due to storage (emmc on HTC 10 vs ufs on others)
4) Now that new phones have so much memory (that is clearly not utilized to full potential), there really should be an update to the android system as a whole to take advantage of it. I don't see any reason for any of these devices (regardless of 4Gb or 6Gb) to unload apps from memory in the scenario of this test. Good thing is that tools/apps exist that can force system to keep apps in memory, if desired, but it is not really a substitute for better native memory management (by android as a whole) that can take advantage of all the ram available.