Hydro now uses a new package to generate the fake laser scan - depthimage_to_laserscane, built by the folks from willow. It calculates directly from the raw depthimage and bypasses the point cloud processing.
Since this changed, it's very likely some behaviour is a little different to what you saw in electric. I've no experience with the electric fake laserscan package, but I suggest you should probably do some experimentation to understand the behaviour of the new package correctly. Some comments:
1) scan height should be independant of the base used - it's the number of scan lines used in the image.
2) We did notice an interesting behaviour once with it at 10. Setting it >1 causes it to scan within an arc out the front of the robot. This can actually create surprising obstacle detection situations resulting in (surprisingly) less robust behaviour. At a distance, it can trigger on obstacles that aren't directly in a horizontal line out the front of the sensor, and up close, it can completely miss these obstacles.
You might like to try setting it at 1 to see if this fixes any spurious behaviour you have. We haven't really done extensive testing to bulletproof the characteristic behaviour of this software yet - it's just a good starting point for people to experiment.