I was just trying to sort out how this feature works, and came up with the following list of issues to be aware of when trying to use it:
when exporting, if a displacement node is listed, then it'll export BOTH Node1 and Node2. If Node2 is blank, it'll use the node number from Node1 in Node2 in the CSV export. This screws things up when re-importing from the CSV file, since the second blank displacement set will overwrite the first displacement set. You need to manually delete the second dummy set on each node.
Second, it can't import on a cnode, it must be a geometry node, so make sure you correct all node number references in the CSV to point to geometry nodes. You'll have to fix the node numbers from within the calc after import
Third, it doesn't seem to read the configuration unit files correctly, and automatically includes a 25.4 conversion factor on the second row of the CSV. adjust the conversion factor as req'd to get the proper values after import.
Fourth, exporting will dump DX1 to DX1, DY1 to DY1, DZ1 to DZ1, etc. as it is in the calc originally. However, on import, Caesar seems to assume that the .disp file is written with Y axis 'up'. If your exported displacements came from a file where Z axis is 'up', then they will be rotated about the X axis upon import. If your source and destination calc are both "Z up" calcs, then you'll need to pre-rotate the displacements counterclockwise about the X axis for them to import correctly.
So long as you take all of that into account, I've managed to get them importing correctly.
Regards,
Gerrit Kirkwood
KBR