I mean, it's difficult to say specifically because there are certain features unique to each Grand Tour. The Giro and Vuelta have much better options for pacing than the Tour, because all the mountains are concentrated in specific areas in France, most of which are around its southern and eastern perimeter, whereas in Italy and Spain you're never more than a stage away from a significant enough range to put a selective stage in. The age of the roads and how they were constructed also changes the character of the race.
And one of the problems is the way that the organisers have reacted to changes within the péloton have solved some issues, but at the expense that often they've created new problems or exacerbated old ones. The lack of audience figures for flat stages compared to mountain stages caused Angelo Zomegnan to bias the Giro way too far in favour of the latter, resulting in a negative impact on the field attracted to race since a Giro-Tour double was off the cards, and costing him his job. Greater control meaning smaller gaps in mountain stages have led to the perception that the gaps created by time trials cannot be countered without changing the balance. ASO decided that since they had already started to simplify the mountain stages, the solution was to reduce the time trial mileage to make the race closer, which in some respects succeeded but resulted in less canvas for interesting pictures to be painted on, because it just reduced the amount of potentially decisive kilometres in the race. Unipublic decided that since the gaps created by their anaemic Unipuerto stages were fairly small (owing to them designing everything around the final 45 minutes of the stage being the only decisive part), the solution was to have more of those stages, because if you're only creating 30 second gaps where you used to be creating 1 minute gaps, you need twice as many mountain stages to create the same gaps as yesteryear.
And yes, nobody likes a race that's settled after the first week, so organisers want to keep things close until late on so that fans don't tune out - but that shouldn't come at the expense of neutering the course. I mean, 18 sprint stages, then a TT, a single MTF and a parade would guarantee that the race came down to the final weekend, but understandably no race organiser has dared go for that kind of reductio ad absurdum situation.
The Grand Départ:
I feel that a good Grand Tour should open with something that will offer up potential to open up a gap but will leave people close enough that in the coming days there is a fight for the opportunity to wear the leader's jersey. It absolutely should not be handed on a silver platter to sprinters who have no intention of trying to defend it (hello Marcel). Wearing the leader's jersey at a Grand Tour is an honour and therefore it should be earned. The two methods I prefer for the opening stage are the prologue, and a slightly hilly circuit type race. I don't mind point to point for stage 1 if it's something like Plumelec 2008 or Landerneau 2021, but something like the
original proposal for the 2013 Giro's opening is ideal - it
might be a sprint, but then the sprinter has earned their right to wear the jersey. The stage might open up gaps but they're not likely to be very large. I like a circuit race as an idea particularly for the Vuelta because it helps attract people for Worlds preparation as well, but also it means you can improve the atmosphere with fans able to see the riders several times. A prologue often has a better atmosphere than starting with a road stage (plus it's easier to police) for a similar reason.
An early means by which to separate the contenders from the pretenders:
The longer every potential contender has something to protect, the more nervous the péloton will be, and the more likely we are to see crashes. Something early on that tells you who doesn't have the form and needs to reappraise goals, and sets the scene, can solve this problem. This is the quintessential role of a Montevergine di Mercogliano. This kind of stage should be around the middle of week 1, I feel. If the race started with a road stage, then it is better for this to be a time trial - preferably a middling sort of length (~25km but possibly shorter, depending on the balance of other stages), again similar to the 2008 and 2021 Tours. The alternative is an MTF which is either a sprint-of-the-elites type that eliminates the pretenders but doesn't separate out the biggest guns, or a shorter and sharper one that is too long for puncheurs but not as endurance focused as later ones - the 5-8km at 8-9% type climbs work well for this. Examples of the former kind in recent memory would be the 2009 Giro to San Martino di Castrozza, 2011 Vuelta to Sierra Nevada by the traditional route, or of course the many Montevergine stages. Examples of the latter would include Arrate in the 2012 Vuelta and Planche des Belles Filles in the 2017 Tour. Yes, here's me praising the use of one of my most hated climbs - because it was used in the perfect way for it to be effective in the overall telling of the race's tale. Once you've settled the natural order of things in the péloton, a lot of the rest can fall into place.
Stages which are worthwhile for the GC - especially those which incentivise racing from distance - on weekends:
These are the days with the biggest potential audience share. Take advantage of that. No putting all the mountain stages, time trials and everything that is most likely to create new fans on days when people are at work, school or college and then put four hours of coverage of a flat stage on a Sunday.
Especially on the penultimate weekend when the race's key period should be taking place. I'm looking at you here, ASO.
Do not put a huge MTF on stage 20 unless the race's balance is otherwise against the one-dimensional grimpeur, and make the queen stage a proper queen stage:
I know it's sorely tempting, but unless you beef up the time trial mileage, there is a very high chance that the best climber in the race already has the jersey before stage 20, and then it's a complete damp squib. See Ventoux 2009 and Zoncolan 2014 for the ultimate examples. A queen stage should see at least 3 monolithic climbs (the Galibier, Bonette, Stelvio, Tourmalet, Gavia, Fauniera type) or if it's not using that type of climb, preferably about 5 major climbs (the Dolomiti, Asturian, Valdôtain type). Something like the 2003 Giro Pontechianale stage, the 2008 Tour Alpe d'Huez stage or the 2011 Tour Galibier stage are good examples of the former, things like the 2009 Tour Le-Grand-Bornand stage, the 2009 Giro Monte Petrano stage, the 2011 Giro Rifugio Gardeccia stage, and the 2015 Giro Aprica stage are good examples of the latter.
Sequencing of mountain stages:
Especially the Tour has a bit of a necessity of doing the mountains in blocks. That's fine, it's geographically necessitated. But if that's the case, pacing is crucial. If you're going to do three in a row, the most difficult mountaintop finish cannot be last, otherwise it will neutralise earlier stages (see the 2009 Vuelta). If you're going to do the short mountain stage, it seems to work best placed directly after the queen stage, for two reasons: 1) the queen stage will be hard enough for domestiques that they will often fall away early on the short mountain stage, and 2) the short mountain stage being short means riders will be less scared of expending energy in the queen stage. And if there are three mountain stages in a row, the first one is likely to see the leaders leave it to the final climb to avoid too much fatigue across the coming days. This is the ideal place to use your Zoncolan, your Anglirú, your Ventoux, your Alpe d'Huez. The queen stage is preferably on the penultimate weekend. The 2011 Giro penultimate weekend is a great example of how to do the back to back mountain stages.
Placement of time trial stages:
Keeping something back for the time trial is too often a killer blow to action on mountain stages. There are exceptions - pleasingly more common now - but typically you need a transitional stage between the mountains and the final time trial if you're making an ITT the final decisive stage, as the Tour often does if the Pyrenees go last, with a flat stage then a time trial somewhere near Bordeaux. If you're going with a late-race ITT, then you need to have a solid length one earlier in the race. You can get away with only one long ITT if it's held mid-race, but if you're going with that approach then it MUST be the kind of time trial that can gain or lose minutes, because this is the only chance for the time trialist to gain significant time. In this case, you need something like 2009 Cinque Terre or 2015 Valdobbiadene. If there is no prologue, then something akin to the 2008 formula, with around a 20-30km ITT in week 1, and a 45-50km ITT on stage 20, would be good. If there is a prologue, then it's best to either go with a mid-race long time trial (probably 40-50k) and a shorter one later in the race (or even, for the Giro at least, on the final day), or a single, but monolithic, mid-race time trial like the ones mentioned above. The Vuelta used to finish off with a 40km+ time trial in Madrid and the race lead would frequently change hands in them - the Giro likewise has had its final day TT change the winner twice in the last ten years (Hesjedal in 2012 and Dumoulin in 2017) but on both of those occasions, it was already felt that they would win and so the ITT was like a coronation not dissimilar to Cadel Evans overhauling Andy Schleck in 2011.
Variety in transitional stages:
It is pretty well-established at this point that I do not like the Team Time Trial, and unless one of the three countries elects a Communist government or undergoes a Communist revolution, the TTT should therefore not be included. However, we do want to balance out the teams that are brought so that we don't have just full on mountain trains running alongside each other. The point of the race is that any stage can be a place to gain or lose time, not have set liaison sections and set timed sections like the Dakar Rally. I suggest that in fact this is one area where the Grand Tour organisers have been excelling themselves recently. The Tour de France in 2021 just published its best set of non-GC stages in years, and looks like they have understood what they need to do with flat stages to make them more important to the overall picture. I would like to see a maximum of four straight up sprint stages. If more stages end up in sprints, then so be it - for example if the organisers prepared a stage that would be susceptible to crosswinds but the weather on the day didn't play ball, or weather conditions meant neutralisations like stage 1 in 2020. Or it rained and the riders just didn't feel like it, like Asti in 2020. No, actually, scratch that last one. This is also where your other rouleur challenges come in - sterrato, hormigón, cobbles, ribinou.
Some genuine endurance stages in the medium mountain and hilly areas:
Not every stage in and of itself is a thriller, but sometimes the role they play is crucial because they are important for the days to come. This is where the super-long intermediate stage comes in. The 2021 Tour's 7th stage, a monument-length hilly stage, is a great example because without the fatigue created by that stage - which itself was good - stage 8 is more tightly controlled than it becomes. The Giro used to be the best at this type of stage, with examples like Pinerolo 2009, L'Aquila 2010 (of course, the strange situation on the road exacerbated things there too), Porto Sant'Elpidio 2012, Ivrea 2013 and San Giorgio del Sannio 2015. Even if they just end up as stages for the baroudeurs, they will have an impact in the long run on the GC mix as well as giving the type of rider that may have had to reappraise earlier goals or that wouldn't otherwise have a chance to take a stage win a chance to do so.
More prestige on the secondary classifications and a better balance of the points system in the GPM, especially in the Tour:
The Tour's GPM was reorganised in 2011 to reflect the "king of the breakaways" element. This had largely been exacerbated by the poor stage designs of 2009-10 where final climbs 60 or 70km from home were giving huge numbers of points. I agree with the theory behind the change they made but I think the imbalance in the points scale is too extreme, such that we're seeing GC riders winning the jersey by accident (Froome in 2015, Pogi both times). That's less a problem if they're winning it over the type of break fodder that is currently picking up the jersey in the Vuelta (thanks to their doing away with the MTF category and giving a lot of points to summits that are obviously not going to impact the race, which is their problem), but when you see something like Carapaz in 2020 or Wout Poels in 2021 finishing just outside the top 10, lighting up the mountain stages... are these not precisely the type of riders who should be targeting and contesting the GPM and making it a worthwhile thing to follow?