Let's say a 4-legged creature has a simple mutation giving an extra pair of legs... who does this creature mate with? He's a 6-legged island in an ocean of 4-legged creatures... And yet it does happen.
I don't believe that there has been such a case. Different fixed numbers of limbs have arisen from different backgrounds in different cases, but apparently not by changing
from earlier fixed numbers.
The easiest case to illustrate is the arthropods. The basic original arthropod plan is essentially a segmented worm with an exoskeleton and a pair of limbs on each segment, and all segments & limbs identical. At this stage, the number of segments, and thus the number of limbs, is not fixed; it varies easily and without any particular consequences. Then the segments & limbs start differentiating and specializing. In all modern arthropods, the first few segments (although not the same number in all lineages) fuse into a head, and the limbs on those segments shrink and reshape into mouth parts. Modern centipedes have mostly stopped at that point, where the number of segments in that first section needs to be right so the head can have the proper structure, but behind that, it doesn't matter exactly how many copies of the generic non-head segments & limbs there are, and they come in a wide range of numbers, varying not just between species but also within them.
But other arthropod lineages tinker with the rest of the body some more. The remaining body segments can fuse into groups like an insect's thorax & abdomen, and separate series of limbs can become specialized for grabbing, walking, swimming, spinning silk, filtering food particles from water, or even breathing, or they can either be suppressed from growing at all, or even grow but be fused in place together to form a new bottom surface of the abdomen. When those differentiations & specializations happen, they happen to specific numbers of segments/limbs, thus locking down the total numbers. So different kinds of arthropods have different numbers of limbs in different sets, but all derived from an earlier state in which the number hadn't been fixed at all yet, rather than by changing from one fixed number to another.
Vertebrates do show some signs of a similar overall path, starting with segmentation and a high degree of repetition of parts and relatively free variation in numbers of copies, followed by differentiation between copies of what was originally the same kind of part, which settles the numbers down to an extent... but just not in our limbs. It shows up in our spines & ribs & associated muscles & nerve structures (more so in fish and snakes than in those of us with big limbs whose development also distorted the torso nearby), and in the repeated arches/loops of our circulatory systems (at least as embryos), and in our pharyngeal/"gill" arches. But no connection between this primitive vertebrate segmentation and limbs is apparent, because all known examples have no limbs at all or just the same few simple arrangements of limbs that seem to have no relationship to the segments of the torso. So again, although we can't tell what made the number what it was in the first place, once that number got set, no more were ever added to it.
That only seems to leave cephalopod tentacles. These are elongations of muscular structures around the mouth, like an elephant's trunk being an elongation of muscles of the nose & upper lip. What starts out as one or two masses (bilateral symmetry) splits into more separate parts soon after elongation begins, like the way the single stump of a growing human embryo's hand splits into five parts which then continue growing independently to form fingers. Sometimes, individual cephalopods have been found in which the number was abnormal because an extra split happened, just like a vertebrate limb with an extra finger/toe. This appears to have minimal effect on tentacle function and essentially none on fertility, but offspring don't inherit it.