Frankly speaking, out of the box terrain feels quite constraining in UE4. My issue through working with heighfields in UE4 was performance. After porting a shader, authored for another well-known engine, into UE4 node-based shader editor, I’ve noticed that terrain render speed in UE4 is considerably lower. This led me to finding of this issue. If you ask me, using dynamic branching while splatting terrain textures in the pixel shader is a must. Surely there is an optimization to discard unused parts of material for specific components, but that is not enough. Taking into account balancing out component size for least draw calls vs more freedom in painting terrain, the latter one always seems to loose.
And then there is this ongoing poor tessellation performance with dynamic shadows.
Right, who cares about tessellated heightfield’s tessellation anyway 
especially where there is POM. But yet again, I can’t get decent cross-layer POM blend without being able to sample landscape weightmaps with an offset. Plus it would be quite cool to have Pixel Depth Offset affecting dynamic shadows.
Physical surfaces for landscapes are also functioning in a very restrictive way.
This feature request has only two votes on it. One of them is mine. How come?
Lack of vertical LOD falloff into the heap of the issues.
I don’t think I have any remarks regarding scultping/painting tools.
The amount of attention landscape is getting seems directly proportional to the number of users requesting it. And so far I have an impression that there are only 3-4 active forum users, that are actually interested in terrain improvement.
To me, landscape system feels like the weakest spot of UE4, sadly.