ECM can remove material at a speed limited by the area of the interelectrode process gap, assuming adequate flushing. That is, with the same current density distributed over twice as much electrode area, you can remove twice the volume of material per unit time. The maximum speed obtainable from a given composition of electrodes and electrolyte can thus be expressed in meters per second, and commonly it’s a few microns per second.
If you want to slice a block of metal in half, one way to do this is to use a wire electrode, which can cut a plane one line at a time. But if you instead use an electrode in the shape of a saw blade, with electrical insulation everywhere except the edge, the electrode surface area is potentially many times larger, and so you can potentially cut much faster.
Taken to the extreme, this logic suggests using a long line of needles or teeth, a few times wider than the process gap, as the cutting cathode; or possibly two converging blades like a “mandolin” cutter. If the tooth edges have a 10:1 slope, so that for each 1 mm of advancement of the blade into the work they get closer together by 0.1 mm, they have 10.05 times the effective surface area of a wire and thus perhaps ought to be able to remove material about 10.05 times faster. Such aspect ratios, or even more extreme ones, should be viable where they are not viable for a saw, because the only mechanical stress experienced by the teeth is the process fluid flowing around them (and probably through them). By the same token, it should be feasible to make them thinner than a saw blade.
However, around the point of each tooth, it would seem that the material removal rate could be a limiting factor, since the point is advancing into the material without this 10:1 “mechanical” advantage. I’m pretty sure there are ways to overcome this and get a locally higher material removal rate; the problem might take care of itself spontaneously (for example by virtue of the high electrical field around a sharp point), or we might be able to advance the tooth into the work nearly parallel to one of its edges, so that only the other edge “cuts” and the MRR required at the point of the tooth is no higher than anywhere else.
At first, when only the points of teeth are advancing into the workpiece, the MRR will be lower than it would be with just a wire, and also at the end when only small bits of workpiece are left in between the valleys of the teeth; so the length of the teeth ought to be small compared to the thickness of the workpiece.
By applying the same principle in three dimensions with pyramid-shaped “teeth”, we ought to be able to rapidly disintegrate blocks of metal. By using many independently movable such teeth, we ought to be able to rapidly shape a surface to any shape we want.