Here' is the description from forum post http://forum.worldwindcentral.com/showthread.php?t=20615 . See the post for the images mentioned.

I discovered a problem while working with high definition elevation data that we have (1 point per meter), confirmed when I checked it also at level 8 provided by NASA.

Height tiles' values at their borders don't necessarily match; that means that the places where height tiles meet must be softened somehow. But the WorldWind has problems showing this, as it seems those values are used at the very border of the tiles, creating Jumps between tiles.
As an example, take a look at the screenshot around an area in the Andes (check the coords from the image), and you'll see that worldwind fails to match the tiles' borders.

Then, if we look at the corresponding 8th level height field files 795_1439.bil and 795_1440.bil, which are horizontally neighbours for their values (say, with ghex2 or any other hexa-viwer), we see that the values are evenly placed across the tile border, so the problem doesn't seam to be in the data.

Of course, you could fix the height tiles information to match at the borders, but that's not how NASA sends that data. Is this a known bug?

I think it's quite an important thing to solve, thought I would hate to see 0.6.0 delayed because of this.