I think there are two issues with the current voting page that any future voting system should address if it intends to be useful over a long term.
First, it needs to avoid too much of a "positive reenforcement" effect on highly voted suggestions. I think we have a lot this in the current voting system because the highest voted results are listed first when new voters come to look at the ideas. This is problematic when voters have a limited number of votes because they are likely to see suggestions they like near the top. Even if they read all the way down the page, they probably won't have any votes left to give to the ideas there (and many won't even look). The ideas that are near the bottom of the list may never get serious consideration on by more than a few dozen very persistent contributors, even if they are better ideas than ones high up in the list.
The Second issue is that old suggestions should not automatically outrank newer ones. If an idea is posted on the day after the new voting system goes live, how will it compare with a slightly better suggestion that gets posted six months later? Some of the voting systems suggested on this thread appear to have effectively zero chance that the a new suggestion will ever overtake one with a long history.
While there are surely many ways that we could avoid those issues, I have one favorite, which has only received a brief discussion:
The idea is to avoid the direct evaluation of suggestions, and instead have voters compare two ideas to each other (the pair would be chosen randomly). When a user comes to vote, they'd be presented with two idea and asked which they'd like to see Toady One work on first. After making their selection, they'll be given another pair of suggestions, and could keep voting for as long as they want. A variation on one of the Condorcet vote counting methods can turn the collection of relative preferences into a ranking of all of the suggestions. The rankings would be shown on another page (not the one where the voting takes place), though the position of the suggestions that were voted on could also be shown immediately after a user's vote is recorded.
This system avoids both of the issues I raised above, because it does not show ideas in any order (avoiding a feedback effect) and allows new ideas to be compared with older ones on an equal basis (avoiding historical inertia). It also allows voters to contribute without requiring them to spend a lot of time trying to evaluate every idea that has ever been suggested. Some users might only enter one or two relative preferences, while others might enter their preferences on dozens or even hundreds of pairings (with N suggestions there will be N*(N-1)/2 pairings, so it's doubtful that anyone will get to all of them). One final advantage is that suggestions that are duplicates of each other could easily be identified by putting that as one option on the voting page. Whereas duplicate entries tend to work against suggestions in most limited votes systems, it won't hurt anything in a comparison system.