Similar debates have run through the social sciences for many decades.  At one extreme you have concepts like positivism, which values some sort of objective provability such as tests of statistical significance.  At the other end you have concepts like post-structuralism, which is big on relativist and subjective evaluation.  Positivism is great but only works when you can isolate variables that are both measurable and significant - which is not always the case, and then sometimes leaves room for debate about the significance of the variable.  Subjective evaluations are difficult to rigourously eliminate or expose bias in.  

You're asking hard questions and expecting simple answers - you're not going to get to have your cake and eat it.

'Fun' is not an easy concept to quantify, although you could probably design a survey that would tell whether an audience enjoyed the game.  Getting a meaningful comparison between two games (or two versions of a game) on an 'enjoyability' metric would be a bit harder.  Ergo, it would be quite hard to design a survey that accurately measured whether a version of the game had improved, and more importantly whether that improvement would lead to better sales.  The best you're likely to get out of this test is a like/didn't like measure that might give you some idea as to whether you are on the right track.  In this case I think the hardest part of evaluating people's responses would be finding ways to measure or eliminate bias from the results.  Designing effective studies to measure subjective things like that is hard and getting it right is where social scientists actually earn their keep.

Essential vs. non-essential complexity is another concept that has complexities of its own.  Information theory can tell us about a minimal space needed to represent information, but relating that to real text of a real game is a much harder achievement and natural language processing hasn't really done an effective job of solving that problem.  So, we're back to editorial taste and opinion about the value of the complexity in a mechanic.  Then you've got differing tastes for simplicity vs. crunch.  There is a lot of room for value judgements about what the purpose is, whether it has been accomplished or accomplished satisfactorily.  Tunnels & Trolls and Harnmaster are both fantasy games, although it could be argued they have different purposes and audiences.  Then there is the question of 'Whose purpose matters - the author or the players?'  Even the attempt to get an objective evaluation out of this question might not be terribly meaningful.

Clearly written and well organised could be tested through comprehension tests of people who have read the rules to some extent.  You might also be able to use applied experimental psychology (human factors) techniques to see things like how many times people have to search through the rulebook to find things.  Whether these measures really tell the whole story is debatable but they might give you some useful insight.  However, setting up a human factors lab in your basement and finding enough guinea pigs for a sample large enough to draw statistical inference from is likely beyond the reach of a typical indie game design shop.  WOTC might have the dosh to do it (although the 5e material suggests they didn't), but you probably don't.

Take a look at anything written by Tim Hartford, The Undercover Economist.  He manages to make books about research methods interesting, and shows many interesting examples of creative ways social scientists have taken to isolate significant variables in a complex, messy social or sociotechnical system.

This brings us back to what Linus Torvalds calls 'good taste' - skilled, practiced judgement from someone with the expertise to evaluate whether something is any good.  Not everybody has it, but it can be learned - there is a saying that goes 'Good judgement comes from experience.  Experience comes from bad judgement.'  Judgement is not necessarily 100% reliable, but it is probably the best tool you have for evaluating the quality or fitness for purpose of a role playing game.  You will be too close to what you've written to evaluate so find some folks whose judgement you trust and ask them what they think.

I see this more this way,

Back Then -- make a character, backstory gets created as we play.

Today -- make a 10 page backstory as you create the character, so that the DM is forced to swallow your claim to the 110 skills that you have mastered so that you can have an Über character.

If the character generation system lets you roll up a character with 110 skills then that's fine or it's an issue with the design of the game mechanics - all the other players could roll up a character with 110 skills using the same system.  If the system doesn't allow 110 skills then redirect the player to use the actual rules for generating their character.

FATE, for example, is pretty free-form in what you can do with a character, but all characters get the same skill pyramid plus 5 aspects and 3 stunts.  You can make up any back story you want but if you don't design trouble into the aspects then you don't get any delicious FATE points from compels or invoke-against.  Mary-sue characters are bad by design.  D&D is even more prescriptive in what skills you get for what class and you can make it even more egalitarian by using point buy for stats.  Traveller only gives you coarse-grained control over your character to begin with.

There is, of course, a social contract here not to be a dick.  Some folks observe this more than others.

I've found DMs pretty much universally appreciating back stories if you put in some decent hooks; I even had the DM requesting them from other players.  You can shake the back stories down and fit them further into the game as you go.  In one case recently the DM took one of the hooks and turned it into a major side-quest for the party.

That's almost certainly what's being done in the OP's examples.  For just about every major commercial font there's a freeware equivalent that's nigh-indistinguishable.  Much of what you're paying for in a commercial font is the kerning and hinting information.

Apologies for the necromancy, but the hinting and kerning information is what makes a font look good.  On display faces at larger point sizes you can work around it by manually kerning and the font rendering OK at size (although this will work a lot better in print than on-screen) but for text being set at 9 or 10 point you absolutely need a font with good kerning and hinting data.  It will look terrible otherwise, and hinting and font design for screen rendering is another whole topic in itself.

Individual fonts aren't all that expensive, even if large collections like Adobe Font Folio can run into thousands.  For a publication you're probably best off shelling out for the fonts you use as there are probably only going to be a few of them.  If you're (for example) a jobbing printer with an in-house art department, or a graphic designer then you may have to shell out for a large font pack.

One little protip that's worth noting is that old versions of Corel Draw go for peanuts on ebay and come with a collection of around 1,000 high quality fonts from Bitstream.  They also used to turn up in bargain bins quite often as well.  Corel Draw is also a fairly decent drawing program and will run just fine on older hardware.  It also comes with a paint program and a whole bunch of other stuff - it's always been good value for money.

And, yes, I used to work in this field at one point.

I have done a few in my time.  At one end, the process ranges from CT/Striker/MT hybrids - with lots of house rules of course, this is Traveller after all.  A more extreme example would be a very hacked about Twilight:2000 1e system that I used for moderns and sci-fi gaming in the first half of the '90s.

This had a hacked about damage system, some tweaks to the character generation system (although it still retained the tradeoff between stats and skills from T2000 1e), plus a variety of house rules.  In all, I ran 4 sci-fi campaigns with it and several moderns games.  At one point I did a cyberpunk hack for it to address some of the flaws in the CP2013 system.

After that I did a brief segue back into using Traveller for sci-fi.  Then I drifted away from role playing in the 2000's and came back about 2015 or 2016.