Semantic model takes 20 - 30 seconds to load on OH 3.3

Update: Rebooting the Raspi (not only the docker container) seemed to have helped (even though I tried this already before). Will continue to investigate.
Update 2: I re-deployed my entire system / docker containers (via this) and this seemed to have solved the problem (also visible in a significantly lower CPU temperature, see below).

Not sure where the problem originally came from (I started with OH 3.1 and did three updates since then), but having a fresh install of 3.4 was obviously a good idea.
Update 3: Problem still persists. Currently on OH 3.4.2.
Update 4: Lack of RAM. D‘uh…