To keep it brief I'll limit myself to 3 each.

Absolute Must Gets

• Role Queue (Pretty much a given)
  
• Hub-Centered Queues (Slightly less important, but could help make Newbie-Friendly Groups for example)
  
• Community Discussions (It should be possible to arrange and run community nights independent of Faceit's team)

Need To Change

• Balancing (If their 12v12 mode is anything to go by, balancing matches is shocking)
  
• Their confusing focus on Casual. (Comp is by far and away a more consistent audience, even if numbers are lower)
  
• In-Game Reporting should be available, possibly implementing some subbing system (Opt-in on queue)

Absolute Must Nots

• Not leaving the community to figure out the system for themselves
  
• Using bizarre community events (WTF was Clan Wars?)
  
• Listen to random content creators for legitimate advice.

The last point touches upon the second part of the question. They should be receptive to all feedback, but understand who exactly is making those points. Whilst an Invite/Prem player may have some ideas on how to organise the queue, are they the ultimate authority on how to run a matchmaking system? If there must be some 'council' or 'illuminati' it would make sense for that sort of thing to involve people who have tried this before or at least have notable experience in organising for competitive tf2. Now that comes with a caveat that even those organisers will be biased and have their own views which may be misaligned with the attitudes with the community, but by bringing enough on you should have something reflective of the competitive scene. Again, all feedback should be given weight regardless of whether its a player, professional or organiser but if you want the most robust arguments on how to manage competitive play, its best to hear from those who have been doing it for years.

.

dqs

they should hire zesty jesus as a consultant, that guys got some solid opinions on tf2

Dunno If this is unpopular but I would actually like faceit 6v6 to have some sort of comms or communication; that way it could be a good alternative to centers. So that high quality games could be played without idoitic flank plays getting rewarded. Maybe you would get a slight point reward for joining a voice chat or something so that people would get talking.

This I think is really important. Although I wouldn't say don't listen to a random content creator but instead: "Don't listen to just one individual".E.g. just because someone has played comp tf2 for many years and is playing prem/invite now doesn't mean that their opinion reflects the entire tf2 comp community.
Best case scenario in my opinion would be to either let the community vote (which requires some good voting options in the first place) or to just draft a bunch of tf2 players from different divisions/continents and ask them/discuss with them what they'd want.
Since I'm pretty certain neither of those things are going to happen I think the easiest thing for them to do would just be to be transparent on what they are trying to implement and then listen to community feedback and make changes but only if they absolutely feel like they must (sometimes I feel like the tf2 community rebels against certain ideas/implementations only to later enjoy them).

just dont have rgl run the 6s queue and not develop a 7s que- woops they did that

Use a server-side anti-cheat plugin instead of a client-based one. Makes stuff smoother for general users.

think about how much trouble could be saved waiting for people to be banned if we just all used the steph anticheat

listening to everyones opinion regardless of div/skill level is the reason shit like the spoon, winger, quickies, loch n load etc arent banned yet
id say that at this point in time making it to at least low prem or invite is required for someone to have a good understanding of what makes the game suck ape dick and what makes it good or at least not as bad to play

In my opinion, having a diverse set of prem to high players give their input on how the system should work - and all players giving their feedback on their experience with prototypes of said system - is how faceit should approach this.

Prem players opinions aren't always gospel, and shouldn't be treated as such - take the removal of medic-scout speed plugin as an example. Majority of the comp community didn't enjoy it, and while it may have been negligible to even enjoyable to some - it's still worth noting prem players suggested that

I've always thought a priority role queue system would be best for comp tf2 matchmaking - for 6s but especially highlander since theres a lot of classes that are less popular. So essentially every player sorts the playable classes from most to least desirable and when theres enough players to form teams, the matchmaker would create two teams of similar ELO and then assign the classes based on priority. If two players on one team had demo as their first priority, one is randomly given demo and the other is given a different class further down their 'priority list'. This way someone is forced to play generally unwanted classes like medic (like traditional pugs) but hopefully they want to play it a little bit more than everyone else - unless of course all players had medic as their lowest priority. hypothetically everyone would be okay with their class and the turnover for games would be fast. FACEIT could also monetize this by allowing premium players to always win those 'tiebreaker' situations (2 demos 1 team)

But theres still the question of whether players should have seperate ELO for each class - feasible for 6s but maybe not for highlander

I mean, yeah and no.

Whilst in my view smart prem players should have some bigger input, ultimately prem makes up a small portion of the competitive player base and in a game that generates no money, who gives a fuck if you´re prem or open. Everyone is here to have fun.

u dont need prem or high players to tell a matchmaking site how matchmaking should work

all that faceit needs to do to live is be marginally better than tf2center (which it failed last time by letting csgo players run perma engi with 70 fov)

the last attempt failed because
1. prioritizing paid premium players as captains, which were virtually never actual tf2 players so they just picked randomly
2. having no class selection, so last person in the server got forced on medic (or the team just didn't have one)