As it says in the Camera node documentation: http://planetside.co.uk/wiki/index.php?title=Camera
"Orthographic: If this radio button is checked the camera uses an orthographic projection. This removes the effect of perspective and can be useful for top down images, for example." So it's similar to what you describe, although it depends on what you mean by "map-style". The rendering style is the same, it's just without the effects of perspective.
The Water Shader does not have any built-in area limitations or masking functions. When applied to the Lake object, it covers the entire object, and is only limited in area due to the size setting of the Lake object itself. When applied to the terrain, a Water Shader will cover everything unless masked by another shader or otherwise limited in its distribution.
I would suggest ignoring the Image Writer Tiff node, it is an older node that is no longer used or necessary.
I'm not sure what you mean by "compress my node network", but you can save parts of your node network into Clip Files by selecting the nodes you want to save and then using the options on the File menu for clip files. Alternatively, if you mean you want to have nodes inside other nodes for purposes of organization, you can do this to a limited extent, for example by putting nodes inside of a Null, however it is a bit cumbersome and is probably most useful only when node networks get really large. Once you get used to working with internal node networks it can get easier and some people do prefer to work that way in general, but Terragen is not currently setup for the smoothest operation with this workflow. Collapsible groups, etc. would help with this.
Painted Shaders can be used for adjustments both large and small. However they are generally intended for quick and easy specific masking or other adjustment tasks. If you have the need to precisely paint a lot of masks it may be better to do so in an external image editor such as Photoshop. Painted Shaders generally use more resources than image-based masks with the same shapes, and they are also generally less precise. If they seem to be working well for your purposes and you are not encountering performance issues then feel free to continue using them. Just keep in mind the use of external image-based masks as an alternative if you encounter performance or accuracy issues.
Your Photoshop question is a bit beyond the scope of what we can provide support for here, but I can tell you there are various ways to accomplish what you're describing. One option would be to simply create your shape in Photoshop, then apply a Blur filter to it with a black background so your white shape "fades to black". The blur radius will generally be defined in pixels, which will give you something like the results you describe.
You can only hook one node at a time to any single input. However you can chain nodes together in various ways to combine the data of several nodes into one for use on the input of another node. For example if you simply feed the output of one node into the input of another, this can combine them in a simplistic way. However some nodes completely cover the results of the previous one, and you also don't get much control this way. An alternative is to use a Merge shader to merge 2 other shader nodes, and you can control how color and displacement are mixed in the settings of the Merge node, including Add, Subtract, etc. So you could, for example, subtract an image map (with a simple shape like a circle, for example) from a Painted Shader to get a circle-shaped hole in your Painted Shader. Or you could combine a Simple Shape Shader with your Image Map Shader, etc.