So I've been thinking about this a lot lately. Obviously my first theory was on the right track (Z/W linkage) but then had to be modified because cinder was only behaving sex-linked in certain females. I admit I was writing pretty casually about this and said some things here and there that are not entirely accurate. My second idea just introduced the idea of crossover suppression but left some unanswered questions, such as why that was happening in some females and not others.
In humans, there are regions of the X and Y chromosomes called pseudoautosomal regions (PARs). Basically these are regions where the X and Y share sequences. During meiosis, the X and Y pair at these regions and crossover occurs. This appears to be necessary for the X and Y to segregate properly into separate sperm (ie, to prevent a sperm from being XY or from lacking a sex chromosome entirely). Pseudoautosomal regions exist in birds and presumably other animals using the ZW system that have heteromorphic sex chromosomes.
Where is this going? Genes that are located outside of the PAR regions exhibit sex-linked inheritance, either because crossover cannot occur at that region or because the gene is only present on the X chromosome and has no locus on the Y. But genes located in the PAR regions don't exhibit sex-linked inheritance. Rather, they are inherited as if they are not on a sex chromosome at all. PARs can vary within the population and there are some PAR variants that exist in only 2-4% of the human population.
Thus, if PARs can vary, that gives a plausible explanation for the bizarre inheritance of cinder. There could be W chromosome variants where cinder is not located within a PAR so cannot cross over, and other W chromosomes where cinder is within a PAR and thus has no problem crossing over with the other allele on the Z. This explains why we have never seen the situation of a female het cinder only producing female cinders: presumably there is no way to get cinder onto a W chromosome that can't cross over with the Z at the cinder locus. This is because any female cinder can trace her W chromosome back to the original female cinder, and those W chromosomes can cross over as evidence by that fact that het cinder daughters of female cinders do no display sex-linked inheritance of cinder. So any het cinder female with cinder on her W had to get it from a homo cinder female, since W is always inherited from the mother.
In the wild, an upper keys corn is geographically isolated from something like a Carolina corn. Heteromorphic sex chromosomes evolve more rapidly than autosomes, so it is not surprising that there would be variation in the structure of the W chromosome in different localities that could prevent crossover at certain regions of the ZW chromosomes.
I'm not saying this scenario is correct, it is just an explanation that is plausible. But the moral of the story is still the same: in cinder projects it is best to use a cinder female or the het cinder daughter of a cinder female, as you will be guaranteed to not have sex-linked inheritance. If your female got her cinder allele from dad, you may get lucky with her W chromosome. Or maybe not.