What I am thinking is this:
Consider there's a conventional wing with spars and ribs and skin all made of metal. Now reflections occur at the metal skin as it is opaque to radar. If now we replace the SKIN ONLY with composite (for example) what the radar now sees is the internal structure of the wing (the spars, ribs, stringers) and they will form many many corner reflectors and I'm reasonably sure the return will be much higher (or am I mistaken?).
So won't a partial composite buildup cause more damage to the rcs than reduction if the internal structure is left unchanged? I can think of one solution right away, using lightweight directed reflectors (foils) internal to the composite skin to the 'hide' the internal metal structure. Is that what they do?
You are mistaken.
You are assuming that composites are inherently 'pass through' and that is not true. A radome is a pass through device by design, meaning its components are deliberately chosen because of molecular make-ups and how they are array together to allow pass through. By the same token, you can chose components by different molecular make-ups and array them differently to completely trap some -- not all -- quantity of any impinging EM signal.
Composites that are specifically designed for low radar observability
DO NOT have that pass through characteristic, but they do have higher 'permeability' and 'permittivity'.
Materials Research - Complex permeability and permittivity variation of radar absorbing materials based on MnZn ferrite in microwave frequencies
The complex dielectric permittivity (ε) and magnetic permeability (µ) of Radar Absorbing Materials (RAM) based on magnetic particles (MnZn ferrite particles) embedded in a dielectric matrix (silicon rubber) have been studied in the frequency range of 2 to 18 GHz. The relative permeability and permittivity of MnZn ferrite-silicon composites for various mass fractions are measured by the transmission/reflection method using a vector network analyzer. The concentration dependence of permittivity and permeability on the evaluated frequency range is analyzed. In a general way, the results show ε' parameter presenting more significant variation among the evaluated parameters (ε", µ", µ'). The comparison of dielectric and magnetic loss tangents (ε"/ε' and µ"/µ', respectively) shows more clearly the variation of both parameters (ε and µ) according to the frequency. It is also observed that higher MnZn ferrite content fractions favor both dielectric and magnetic loss tangents.
What it means is that the composite material will allow penetration of its surface, but not pass through the entire thickness layer, thereby in a manner of speech, the composite material essentially 'traps' the impinging radar signals.
So keep in mind...
Penetration =/= Pass through.
No composites is ever perfect. The composite material will always reflect some measure and permit penetration some measure. So because of this imperfection, radome design and construction is critical to reduce what is called 'radome aberration' where the radome can and will produce false targets.
But there is an interesting twist...Just because you intend your composite material to allow penetration but not pass through, that does
NOT mean the design is any easier than the formula that was designed for pass through. The formula for high penetration but not pass through is just as difficult to create and manufacture as the formula for complete pass through (radome).
The reason came from the radome itself, as in a bad formula and poor quality control manufactured radome composite material that produces false targets.
If a flawed radome composite material can produces false targets at one or more specific sweep angles, a composite material that was intended for high penetration but no pass through can be so badly formulated and manufactured that this formula can produces high reflectivity at certain surface locations but not other, thereby defeating the purpose of 'stealth' in the first place.