Not so much -- I never seen 'em used for induction heating.
Why? No cooling! The poor thing will melt, from the inside out, in under a minute (and that's at low power levels of several kW).
Square and rectangular tubing is used a lot, though: but not because of losses. It's easier to fabricate. You can chop a piece into segments by making miter cuts (sealed with silver solder). It bends easily without kinking, even for fairly tight inside corners.
Indeed, the losses are higher. Much higher! For a coil with dimensions as pictured above, there would be extremely high eddy currents induced in the coil itself (the eddy currents flow across and around the conductor, while the load current flows lengthwise). This forces the load current towards the inside corners (path of least impedance). Indeed, the effect is so strong that the outside corners will carry opposing currents (going backwards!), if not as intense as the inside corners' current.
For any application where low losses are required (RF tank coils, most induction heating apps), round tubing is best, spaced with a pitch twice the tubing OD. This reduces the eddy currents, caused by the turns being so close together, without reducing the overall inductance too much (because it is stretched, relative to having pitch as close to diameter as possible).
To put concrete numbers on it: the pictured coils might have a Q in the 20-50 range, depending on shape and frequency. A loosely spaced coil of round wire/tubing will have Q ~ 100-200, all else being equal. That's a big deal when you're resonating over 100kVAR!
They do work nicely for cored inductors, and lower power levels. All the big name manufacturers have lines of inductors made this way: for example, Vishay IHLP series, lots of Bourns and Cooper (now Eaton) high current parts, Coilcraft high-current parts, etc.
And of them, the powdered iron types are actually molded around the winding, so the core material completely fills the gaps and crevices between turns. You can't beat the amount of inductance available this way -- these things have dense ratings -- lots of inductance and amps in a small package.
The Q factor improves about proportional to permeability, so they don't get terribly hot in the process, either. Add a permeability of 10, and Q shoots up from 20 to 200. In practice, it's worse than that, because of geometry factors and core losses. Typical Q for this type is 50-200: quite reasonable for a power conversion application that only needs to circulate 100s of VARs.
And what is square/rectangular wire most ideal for? The applications that don't need to worry about skin and proximity effect: mains frequency transformers! The high winding factor (most copper area in the winding space; packing factor) allows a good ~20% improvement in current capacity, over round wire, for a transformer of identical construction (same wire width, same number of turns).
Tim