Sounds like more complication than needed to actually drill it down to chromosomal separations. While there are some things that are complex interactions where it's relevent whether it's different bits of the same chromosomes or on different chromosomes (not forgetting the 'proteome' soup, directly from your mother's early development, and thus a result of how your grandmother was affected by the likes of hunger/stress at or around the time she conceived your mother), it might as well just be bare phenotype choices.
So I just say that where you currently have "hair colour: auburn" and "eye colour: emerald green", and the rest, defined in a current body instantiation, you now have "hair colour: auburn|blonde" and "eye colour: emerald green|brown" in a "quality: mother's contribution|father's contribution" format, and then either randomly (or by preference, arbitrary or by global rule) express just one of the choices. But, as and when they become a parent, either of the choices could be selected to be their parental contribution.
Basically, this doubles the "qualities" definition, in the relevant creature's data-structure (a small increase, barely significant on top of the other data an entity holds and changes, day-to-day), and adds a bit of choosing code, and makes redheads (or whatever is an equivalent 'recessive' quality) rare, but occasionally springs a 'sport', even in the most carefully husbanded population.
If we're going for full chromosomal information, however, there's going to be opportunities for junk DNA areas to allow genetic 'tricks'. Normally unexpressed areas can perhaps contain historically-justified developmental features in an "inactive" form, you'd be looking at random mutations changing those features and/or whether they're expressed, so occasionally a dwarf will be born with webbed fingers, or something. Here you have an additional load of "historic" genetic forms (webbed fingers, quadruped form rather than biped-with-arms one, gills, feathers, whatever is appropriate) with an associated 'chance' weighting and (assuming that the chances are weighted so that 'chance of being viable' is already a given end result)
But that interesting proposition probably needs some examining of the current biotemplates to populate all biped homonid genomes with quadraped-like qualities (and/or prehensile tails) in these 'hidden' spaces, either explicitly or through some sort of procedural/worldgen-initiated process.
(So, in an hypothetical example, who'd have thunk? It turns out that Homo Nano (dwarfkind) is more closely related to Homo Khalkina (kobolds) than Homo Hominis (yer bog-standard human) and so are more likely to 'throwback' to all-over body hair than to become tall and beardless abominations (and far too linked to close relatives Homo Arborealis/elven, anyway).... Or whatever.)
 Chimeric possibilities, aside. I'm not sure whether we should complicate things so that some dwarves express different coloured eyes, or different coloured hair on different parts of the scalp (or more generally across the body). And you know there are examples that during paternity-related blood tests, mothers have been found to have a different set of genes in their blood (those cells that contain DNA) from their reproductive systems, and have appeared, at first sight, to not be the biological mother of the child that they are inarguably the actual and known mother of.
 Perhaps if they've got some mystical aspect to their particular make-up by default, each eye would be of either parents' colours. Which might not be noticeable, if both gave the same 'token', of course.
 Although that's common enough already, and dwarfish animals have really wierd colour combinations to different body-parts, already, so the genetic features that create (in females only, as I recall) "tortoiseshell" cats are pretty much already overshadowed.
 Although being born with gills but no lungs would cause problems for a child, unless you can get the mother to stay in the partially flooded training pool, until the infant becomes an independent youngster who can aquatically operate on their own without risk of "air drowning" while in maternal care. However, an amphibious form that's capable of both environments (may require occasional wettings) would be a more likely throwback form, and make for an interesting unit to have. I wonder if you can make a flooded barracks for them to train in... Hmmm...