The small amplitude noise works well if it's applied several times with frequent pitfills in between. I usually get best results with Absolute Magnitude noise and a touch of Percentage noise for high areas. The pitfill tends to drown small noise wigglies, but if you keep adding noise on top of the flat areas generated by pitfill, you eventually wind up with a lot of small variations. This also happens to be one of those weird exceptions where it is better to start in low-res resample up and build smaller and smaller details.
One thing I liked from Bryce was Slope Noise. I think I can do workaround for that effect in Wilbur. Build your desired initial hf, and save it. Create a greyscale slopemap in textures. Apply the texture as a selection. You may need to invert the selection cause I think Wilbur shows flat surfaces as white. Save the selection with a memorable name like slopeselect.bmp. Reload your saved hf, load your shiny new selection mask, and just add noise. But I digress.
Deterministic flow algorithms almost always fail in the same way. I personally prefer Wilbur's Incise Flow to the Flow Accumulation tool that comes with ArcGIS. That thing is uuuhg-lee! The biggest problem is rivers go straight right where real rivers get all wavy: on the flats. The problem is a lack of data on which to determine flow direction. Up in the hills, eroding a place down a few inches here, depositing things up a few inches there doesn't make much of a difference when the river is surrounded by steep slopes hundreds of feet high. If the river is in an area where the surface is flat, a bit of deposition will turn the stream drastically away and a bit of erosion will pull the river in. If you don't simulate realistic erosion in excruciating detail with a lot of iterations you don't get that effect. I've seen sims like this, they usually cover a short reach(tens to hundreds of meters maybe, not more than a few miles) in very high resolution. They're also time-consuming.
Many, very subtle iterations of Incise Flow and Precipiton Erosion with frequent small amplitude noise applications might get you a decent approximation, but that would be pretty slow, too.
It might be nice to have a tool that determines limits of a river meander area based on a threshold for local elevation change. Say a selection of all areas within a user-selected height above a nearby part of an existing river selection. Then you could draw new river meanders inside the selection boundaries.