As the whole of the outer casing is in effect a multi-directional bumper for detecting collisions, the 2 stabilising springs (or "feder" - "feather", as Karcher now call them) seem only to stop that shell bobbing around excessively. Without those "feathers", the bump-detection is too sensitive and triggers too often (on carpet especially, less so on smooth, hard flooring).
But I think this OP's problem might be a failed main board: He has no drive to the wheels, yet the sequence of LED's and warm batteries suggest power is being supplied as if the robot is otherwise functioning normally. If it were failed wheel motors, it would be very unlikely that both would fail simultaneously, and the robot would likely drive in a circle? Power is also applied to the wheel motors to apply pressure against the charging contacts when docked, where here the robot is just rolling off which also supports this view. Just a random theory; I've never heard or read of any failure such as this before.
2 x Karcher RC3000 (2005 & 2012), Trilobite ZA1 (2005), Navibot 8855 (2010)