Type 43 and Type 23 were probably the first to have design elements to reduce RCS to some degree.
Even so, I'm doubtful how much of the sloped panel craze in modern ships is really just fashion rather than any serious attempt at RCS and whether RCS really matters on naval vessels.
Well, if you're saying that the Zumwalt hullform probably is not the wave (ha) of the future, I agree. But "craze" and "fashion" are way off-base.
Let's remember that all radar (and sonar, for that matter) techniques are probabilistic things. We live in an EM-noisy environment, and detection is based on the agglomeration of thousands of individual pulses. A low radar signature is not some sort of magic scaling factor that reduces detection range by a factor of X in a universally consistent and repeatable way.
Suppose you've got a MEKO-style inward-outward-inward-outward-creased superstructure. And ships roll, the observer says, and so therefore who cares! But the ship will roll those reflection spikes at a lazy, moderately variable rate. That roll rate will form a convolution with the pulse repetition frequency of the interrogating radar. Maybe the radar will catch a tenth as many pulses as it would from a ship with a lattice mainmast chock full of corner reflectors (which reflect an incident beam from any direction back at its source), and these pulses will come in a relatively chaotic distribution. It will look very much like noise, and for a radar to detect the target anyway, a much higher false alarm rate will result. Or you'll have to use much more complex processing algorithms, which might be too expensive for a ship radar, or too bulky for a missile one.
And, frankly, it's pretty cheap. Just because relatively enclosed, flat-surfaced designs aren't amazing proof against radar detection, putting a mild steel box around your RHIB is not a huge ship impact. It's good return on modest investment.