The answer to all this can readily be found in the pages of Charles F. A. Fyfe, Steamship Coefficients, Speeds and Powers, 2nd ed. (London: E & FN Spon, 1920) (available on Archive.org) which tells you everything you need to know to choose ship coefficients and dimensions to meet a given set of requirements. Computers have made it easier today, and we have more data to work from, but the basic principles are the same.
The S.S. United States aside, no liner I know of ever had a service speed whose numerical value in knots significantly exceeded the square root of her LWL. That is, for liners V/SQRT[L] < 1.
Contrary to what ADeblois appears to assume, LWL does not inherently impose any limit on the speed of a ship. If it desired to design for V/SQRT[L] = 50, it's entirely possible. But there's a price, particularly in power and hence also in fuel consumption. Again leaving the special case of the S.S. United States aside, the tradeoffs always favored V/SQRT[L] < 1. For the Olympic class, V/SQRT[L] < 0.8, permitting a considerably fuller form than the Lusitanias. That brought all kinds of advantages from the standpoint of design and economy.