Hey, thanks for the answer. My project is basically 99% code, so SetBase could work but would be really inefficient for the way I am doing things.
I am working on a theoretically infinite world which is constructed around the player in form of procedurally generated meshes. Since it is a Top-down game I am generating a fixed amount of chunks around the player - just enough to fill the screen.
When the character e.g. reaches the next chunk in forward direction, I take all the dynamic mesh components from behind him and place them at the front of the currently visible chunks and rebuild the meshes for just those components. This way a character can walk around infinitely.
The problem is - the server needs to have the visible chunks from all other players too (for networked collision physics), clients only care about their own chunks - it doesn’t matter if other players, which aren’t in range of the own chunks - and thus can’t be seen, fall through the floor on the client side. If the server now moves the chunks around to rebuild them at other places, it will move the pawns that stood on those chunks with them - since the server has the authority over movement physics.
I didn’t really think about the SetBase() method - so thanks for the suggestion, but my structs, which manage the chunk movement don’t know which character is currently standing on which chunk. I could iterate over all the characters for each chunk to be moved and do all of that, but like I said it would be inefficient.
At first I experimented with just having the dynamic mesh not have any collision at all and just place a huge collision box at the floor of the map - since the idea was to only have totally flat tiles (but I am not really sure if I want them to stay flat anymore), but huge collision boxes have extreme precision issues - the larger the box gets, the more the characters stutter while walking over them (weird thing is, a huge BSP placed in the level didn’t have those issues).