Why just one head on an actuator arm? Put as many heads as will physically fit in a series, to put different heads on different tracks simultaneously. If you interlace the data right, you'll get throughput increased by a factor of X, where X is the number of heads on an arm. You can also put them in a grid rather than a line, with successive rows offset to each other, to get more in-between tracks. This will work a lot better (or at all) if the range of head motion is perfectly along a radius of the platter (and so would be its row(s) of heads).