Without seeing semper Fi it's hard to provide detailed feedback on what needs to change for HOI4 that's sort of immediate gameplay stuff (obviously the defence Ai is broken, but they're planning on fixing that for example).
So a few things that are more 'systems' oriented. An actual globe (either that automatically does a projection, or just doesn't use one at all) would be nice. That's easier to do in DX11 than anything previous but that's a nightmare for older hardware compatibility.
Sort of by extension, a cleaned up graphics engine might be nice, the performance of the current one is ok, but they could use more threads for example and some more fleshed out Animation systems.
Sound: Star trek online does this neat thing where it integrates with itunes etc. Whatever one may think of the rest of the game this isn't a bad idea. HOI, vicky etc are all LONG games, which is good, but getting a diverse set of music for that isn't really viable. A better music system so I can play tracks from other stuff or even other genres, maybe I want to listen to Bruce Springsteen when playing a Vietnam scenario... which takes us too....
Timeline. I would, if it were entirely up to me, move HOI to start in 1910, and then basically have two paths, one for central powers win WW1, one for allies win WW1 that both lead to a variant of WW2, and then let the game go wildly ahistorical and push a WW3 that's between the two surviving factions of WW2, assuming they exist. This could be neatly done with two escalatable lower intensity wars (korea and veitnamn naturally). Granted the latter part may be too much. IMO ww1 has more in common with WW2 ai systems than the franco-prussian war or the US civil war. However the latter 'low intensity' stuff is a bit different, I think there's a lot of interesting opportunities there. In the 50's you could still have devastated germany and japan unable to help the allies particularly effectively, but by the 60's they're much more rebuilt, but then have you just traded territory for time? The war in china should play out a bit better, especially once japan falls (if it falls naturally). Starting in 1910 means Vicky would have to end in 1910, but I'd let it start in 1800 with the player fighting the Napoleonic wars, and then the vicky period is the post war century of change (to either regain territory lost to france, or clean up the mess france created). A possible system here is that if a 'parent' state collapses the 'puppet' states fall into dissarray, so for example no matter how much territory japan has in china (even all of it), if mainland japan collaspes the the whole area errupts in various regional fighting for control (the same could apply to a germany that loses control of germany proper but hold territory in the east), these 'unruly' state like entities would then need to be brought back under control. That would require an system that allows say any chinese faction to (re)form china, or any german faction to reform germany (and would easily handle the DDR/DFR split).
Economics and production: The Vicky system is good, but a bit too complex for a wargame. I'd strip it down to maybe 10 manufactured goods, and 10 resources, if that many (right now it's basically 4 resources and 4 manufactured goods, if you count money). Same sort of thing with the pop system. Democracies should be very limited in how large their regular armies are, and everything else built must be made as a reserve unit (which would then force the 'nato' allies to demobalize after ww2, which sets up the balance for WW3). If you brought in a (simplified) version of the vicky pop system, notably for national vs non national pops you could introduce some interesting mechanics. For example, french national armies working for the nazis gets bonuses/penalties when say, defending on german territory (fight to the last man) or the like. They'd present an interesting strategic choice, and of course the british empire build colonial vs national units. Vicky has a more continuos production model, I like that better. Lets say I'm building 600 planes a year (100 to an air 'wing'). In the HOI3 model I get six air units at the end of one year, under the vicky model I get one every two months. The latter is generally preferable, especially to help out a desperate AI/player (the HOI3 model favours early buildup too much). If one were to do that, units should have a production materials cost (like vicky) and an assembly time (again kinda like vicky) but the assmbly time for things like tanks divisions, air wings would be very low, but battleships and aircraft carriers high (but the component, 'steamers' or just 'shipyard time' would be easy to make and is used for all ship types). You'd still have the somewhat exploitable build 20 ships at once, but to have the material for that you'd needed to have been building ships for a long time.
There are a few other things.
Better AI coordination with the player (and possibly with itself). I know they're working on this in semper fi, but there's always more that can be done in this area. For example allowing the player to take over allied nations (say as the UK being able to manage what happens in iraq, yemen and maybe even canada/australia once the war starts), and then you're worried about domestic production and politics systems several places but not immediately, and not necessarily if you want to pass it off to someone else.
Tech trading/alliance research. I can't forsee a situation in WW2 where if canada wanted the lastest tank/fighter/infantry weapon designs from the UK that it would have been refused, or the US once it joined the war. Actual puppet states (iraq, egypt etc.) yes, but not the rest of them. A better, more coordinated research system for an alliance, but of course it's penalized for having too many cooks so to speak (less efficient, but more of it). Maybe allow the alliance leader to 'let in' other members for joint research, but then the more countries you have the less efficient it is (to a point naturally). One would want an allied 'experience' pool too (say the alliance gets 50% of all experience earned, and then you use the higher of your own, or the alliance knowledge pool)
A system I would love to build but might be too much: If you could separate the manpower and material strength of units. Captured units give up their unit equipment (and you could sell/give equipment) sort of like a trade good. Upgrades would be producing some or all of a new unit and swapping the manpower (could happen automatically), but this way upgrading would generate practical experience, and old units could be sold/given etc. If you used a 'vicky' style model the AI would need to buy military units (with cash) either domestically or overseas and it buys the best units it can afford, and is especially pleased if it's given units which it doesn't even have to pay for. Granted you're building a fairly complex new trading system to support it, I'm thinking it's an n^2 problem. Basically you allow trade with someone, and then when going to buy units you search through whomever has allowed trade with you to see what units are available to buy, or assemble your own. I'd have to prototype it and see if it's actually fun or just confusing but I think it would open up a lot of neat stuff that you can't do right now.
Hexes: In a game where the map is known in advance, and some of the pathing can be computed in advance (or otherwise given hidden waypoints and so on to get stuff to go in sensible paths) hexes don't get you much. In a game with a random map a regular geometry that's easy to do a traversal on is preferable. Tolerable for Civ, bad for HOI. I suppose I could buy hexes for large interior areas (sibera, the US, northern canada, brazil, sahara) for simplicity, but there's not much point in doing that if you're building more general provinces elsewhere. If they wanted to throw on a random map random war generator it might be cool. Very costly to actually do properly though.
I'm not sure how well it would play out, but more subdivisions on the naval map (but then a higher detection chance when you're actually on something) might encourage a more dispersed/less overly concentrated force setup.
Edit: Supply system that handles allies well. (and by extension if you let an AI control another merge their supply systems). Fighting in belgium in europe as the UK or persia as the UK is a nightmare to keep things supplies.
An 'occupied by alliance' might be preferable to occupied by specific country. So control is exercised by the alliance leader, even if the one actually occupying it isn't the alliance leader (So canada can take the lead on the invasion of france but then the UK can run most of the occupation, and you don't end up with parallels supply lines for UK/CAN/US).