Mapgen oddities amplified by mapcleaner

[SwissalpS]

I'm relaying this for Huhhila. I'm not sure I understand the issue correctly. Please read the comment to:
https://blockexchange.minetest.ch/schema/Huhhila/mapclean_y2026_ob2_liquid_mapgen_interactions

My own observation:
I had a digout in a pattern spelling "Luanti". Only the top was protected and the depth was immense.
After mapcleaner the top part is still there (as expected). When dropping down to where the mapcleaner
restored chunks I noticed something strange. At first it seemed like I had reached the top of a restored
chunk but after a few seconds my avatar was engulfed in stone. After returning I used postool to show
block and chunk borders. The avatar had to move down into the chunk's second block from the top to
activate mapgen and restore the dugout area.
I was expecting to be able to run along the top of the chunk that needed to be regenerated, to trigger
mapgen but apparently this only happens when a player goes into the second block of the chunk
that needs regenerating.

I'm suspecting the offset by 2 change mentioned in Huhhila's comment has something to do with this.

[SwissalpS]

The offset may also be causing rotated water to be reset when jumped (jumpdrive) near a mapcleaned area.

[SwissalpS]

Huhhila adds:
I was thinking that the "- 2" at GetMapblockBoundsFromChunk should be restored and GetChunkPosFromMapblock
should utilize floorOfDivision((x + 2), 5) and so on. And new unit tests for those two methods similarly to that TestGetMapblockPosFromNode... -32,-32,-32 should be in same mapchunk as 47,47,47

[SwissalpS]

In the past month several jumpdrive ships were either partially or fully devoured by the mapgen offset kicking in.

Just today a player was digging manually downward, fell, died, bones were placed but couldn't be found.
I Verified in logs that mapgen kicks in right after [bones] reports placing bones-node and thus clearing them but leaving the HUD marker.