I was quite far off in the previous sessions, once I finally figured the general idea and had a stepcount closer to valycian's record. I can then optimize it gradually.
Actually, the path is very reasonable but sometimes we might just not be able to pick up all the small cuts altogether.
Finally solved the last one of the series. Even though I know where to approach first after having tried the previous ones, it is just difficult for me to do it properly. Really clean execution by mathmasterzach.