If a character specalizes in a skill it is possible using the 3.x systems to get very high bonuses.
But here's the rub.
Is that a problem?
what is "reasonable"?
If a character devotes 90% of their resources to be good at skill X how much better should they be at that skill over a character who devotes 10% of their resources on the same skill? How much should level play into it? What if the 10% player is 10 levels higher then the 90% player?
IMO the 2e skill system puts too much emphasis on level and doesn't allow enough differentiation based on training and skill. I don't think anyone (that's still playing pathfinder) has a problem with how high a particular character's skill bonus can get. It's only a problem when a player finds a way to abuse the system that trivializes some aspect of the game.
"no one else even needs to bother with this skill check because only one of us needs to pass it and bob gets a +50 when the most anyone else gets is a +14"
It also doesn't help that people overlook the fact that in the above scenario, bob is substantially worse at everything else then anyone else in the group. When the skill system gets complained about, I have to wonder what that person's experience was and what their expectation is. I get the impression that some people feel that they should be allowed to roll for every check and have a 70% success rate when they do, regardless of their character's investment in skills. Sometimes for no justification other then because their character is "high level".
What are your expectations?
It's hard for me to say if your proposed changes would work for you or not since I'm not sure what "feels right" to you.
One of the main points raised was that the discrepancy caused there to be times when the whole group failed an activity (one failed Stealth check spoils a covert mission) or the rest of the group sits around while one character does things alone (solo scouting, solo lock-picking, having one face character).
The game needs to be looked at as a whole to determine if the fact that those scenarios happen is a bad thing. For example, did the stealth mission fail because the party didn't think things through or because there were no other options?
If it's more because of the latter then the former then yes that is a design problem. Magic is a thing in the world and spells like silence and invisibility (which are relatively low level) can cover characters who don't have the ability to participate on skills alone.
As for everyone sitting back while one character makes a check. I see aid another get used all the time at the table. This allows characters with some skill to still participate in a meaningful way. When I'm DMing I encourage this and have house ruled that for every 10 points you beat the DC 10 by you grant the person an additional +2. In this way a character who is decent at a skill but isn't "the best" still feels like they are contributing in a way where their skill matters.
In the case of solo scouting, dropping the veil of secrecy and using short hand makes this not as bad. In the game I'm currently involved in when someone says they want to scout the DM has them roll only one or two stealth checks to see if they are able to scout without being noticed. Then he reveals the map and gives descriptions to everyone at the table with the assumption that the scout told everyone everything they saw. Everyone stays engaged while still allowing the scout to be useful.
4th edition's skill challenges are an interesting solution to the problem but not one that I personally liked. Since, if my character happened to bad at any of the checks I was allowed to attempt, I was much more likely to hinder the group. What's more even if I knew I was going to be bad in the situation I wasn't allowed to sit back and not screw things up.
While it might be annoying to not be able to participate in the action. It's terrible being forced to participate knowing your participation will hurt the group.
Ignore what "feels right" and focus on how it would function.
Ok, well you would have to greatly alter how many skill points characters get. If I ignore the following skills for a moment (linguistics, perform, craft, profession, knowledge) there are a total of 21 skills. With your system a character would need 84 skill points in order to cap out every single skill. This might seem like a lot but you'll find some characters will hit this cap really fast and others would effectively never get there.
A rogue with a 14 intelligence has 90 skill points at 9th level.
A Bard with a 16 intelligence has 90 skill points at 10th level.
A Wizard with a 20 intelligence has 84 skill points at 12th level.
A cleric with a 14 intelligence has 84 skill points at 21st level.
A Barbarian with a 10 intelligence has 84 skill points at 21st level.
A Sorcerer with a 12 intelligence has 84 skill points at 28th level.
As a result, after a certain point some characters are literally legendary at every single skill in the game and others are barely experts in a handful of skills. This seems like a horrible situation since you are heavily penalizing anyone that intelligence isn't a primary stat for and they don't get a lot of skill points at each level. While simultaneously forcing other classes to pick their 5th profession to be legendary at because they literally, don't have anything else to spend their points on.
So if you were to backport it, this is how I think you should do it:
Untrained: 1/2 level
Trained: level
Expert: level+3
Master: level+6
Legendary: level+9
In terms of equivalency, trained gives you the same benefit as if you'd invested max ranks, expert would be max ranks plus class skill, master is max ranks plus class skill plus skill focus, and legendary is max ranks plus class skill plus skill focus at 10th level and above.
As LordKailas points out it's much easier to get bigger modifiers in 1E than in 2E. As a comparison, a typical low-cost magical item in 2E to boost a skill will grant a +2 bonus, while a similar item in 1E will grant a +5 bonus. So we'd need bigger bonuses to be consist with 1E's skill math. Tripling them happens to hit similar thresholds to existing PF1 game mechanics, so I think that would work best. This isn't really possible with untrained (where -12 would just be comically bad) so it makes a bit more sense to go with +1/2 level there.
Moving along, instead of skill ranks you get a number of trained skills. Unlike in PF2, you can be an expert in a skill at 1st level. Any skill in which you are trained that is a class skill is automatically promoted to expert. In addition, you may choose to promote a non-class skill to expert proficiency by spending twice as many skill training points on it (so you could have one expert skill instead of two trained skills). If you multiclass, you use the higher number of trained skills between your classes. A Fighter/Rogue would have 8+int trained skills, same as a single-class rogue, regardless of how many levels are in each respective class.
I'm not quite sure how you'd handle master and legendary proficiency. The skill focus feat would be one way, but I think you'd want there to be other ways. As well, these rules do make it a bit too easy to grab lots of skills with 1-level dips so you'd probably want master and legendary skill advancements to reward taking lots of levels in high-skill classes. I'm not completely sold on how to do that, and once again LordKailos points out the potential pitfalls there, but I think it could be done.
As a result, after a certain point some characters are literally legendary at every single skill in the game and others are barely experts in a handful of skills. This seems like a horrible situation since you are heavily penalizing anyone that intelligence isn't a primary stat for and they don't get a lot of skill points at each level. While simultaneously forcing other classes to pick their 5th profession to be legendary at because they literally, don't have anything else to spend their points on.Okay, you're right, that does seem excessive.
yeah, thinking about it further. Maybe it would make more sense to adopt the non-weapon proficiency system from 2nd edition D&D. You got a bunch at character creation (like 4+int) and then you would gain an additional one at a set rate based on your class, typically 1 every 3 levels or so.
I realize now after writing it out that It's very similar to what Dasrak suggested and as was stated may suffer from multi-classing abuse.