We can imagine or even create experimentally a very short light pulse. Probably the most straightforwards way to model how it spreads is to use an affine parameterization to describe how it spreads. If you have a monochromatic beam, the affine parameterization will be related to the wavelength of the emitted light and will mark out equal intervals along the light beam.
A rest frame can still be defined for such a pulse, it's the unique frame in which the light remains monochromatic as it expands. In a non-rest frame, you'll see some of the light red-shifted and some blue-shifted, so it won't be monochromatic through the whole sphere.
I believe that the expansion will be spherical only when it's monochromatic, due to the affine parameterization argument - we can say that at some particular instant, the pulse wavefront will be N wavelengths away from the origin.
While it's simplest to understand if the light pulse only contains one frequency in the rest frame, I don't think it's vital to the argument.