1. I run memtest for minimum of 12 hours .... usually 24.
2. I don't see the logic in starting the tests 1 stick at a time ... been building PCs since early 90s and have had less than a handful of bad sticks. So let's be generous and say 10% chance of failure per build (5% on each stick).
a) Testing 2/2 sticks at a time w/ 12 hour tests means 120 hours on 10 builds.
b) With that assumed 10, one build will be an issue... so will spend an extra 12-24 hours testing those 2 sticks one at a time ... Thats 132 to 144 hours
Doing 1 stick at a time, that's 240 hours
3. I recommend adding all the RAM ya want from the getgo .... 2 x 8 GB is much better then 2 x 4GB now and 2 x 4GB later. First, having 4 sticks to manage over 2 oft means lower CPU OCs ...2nd, even is same brand same model is "iffy".
a) RAM from separate packages is not guaranteed to manufacturer to work together, s if it doesn't no recourse.
b) Sometimes your may find yaself where same make and model aren't available. Once was asked to add 2 more sticks to a Corsair RAM equipped box and the only RAM with matchin specs was GSkills. My son's box (Mushkin) also was felling its age and same specs so I bought two packages. The Gskills would not play well together in ether box. I tried th 4 Gskills in the users box and they worked fine. I added the users original Corrsairs to my sons Mushkin equipped bix and also worked fine.
c) Saw this again with DDR3 on Z87 .... user had Corsair Vegeance Pro DDR3-2400 2 x 8Gb and wanted to go to 4 x 8GB. Purchased another pair but of course was later version. The initial version had 10-12-12-28 timings ... the 2nd pair was after version 4.51 and it has 10-12-12-31. Even dropping the timings to the 10-12-12-31 didn't work. After some research found out that Corsair stopped using the premium Hynix modules after 4.51 and switched to another source. Send the new Corsair's back, purchased a set of Mushkin 2400 (10-12-12-28) with Hynix modules and they worked just fine.
In short, nothing to be worried about running memtest, just gotta be careful on how you interpret results and do intensive research when ya find an incompatibility.