Getting +5 for every 'win' has been around forever. It wasn't a problem, when at least half the puzzles you were fed were challenging. It became a BIG problem when many people (those with higher ratings, at least) started getting fed long strings of puzzles 1000 points lower than their skill level, that we can get right and run our ratings even higher.
If the eventual goal is to drag the puzzle rating distribution into rough alignment with game (e.g., blitz) ratings, then why not put a pause on puzzles for one day, announce that you are going to map the ratings of all existing puzzle players and all existing puzzles downward to resemble the other distribution, and let everyone loose again.
The "mixed situation" that Martin has mentioned is, as days go by, expanding and extending the long tail of high puzzle ratings (only for those who continue to play, vs. the wise people who are letting the bugs get worked out). It seems virtually certain that this is WORSENING the original problem!
Yeah I don't think +5 is a bad thing. The problem is that at the extremities people keep getting the same puzzles, of which they remember the solutions, and hence these 2^16 puzzle ratings have happened. One guy still has a 65540 rating on the live leaderboard.
To fix it, just introduce more puzzles to the set. 500k puzzles at chess.com compared to 2.5M puzzles at lichess. Just use a computer algorithim to make puzzles like they do. They have fine tuned it to be from human games, so its actually quite good now.
The one thing that chess.com does better than lichess in the puzzles section is the ability to replay the puzzle that you have just completed, and also look it up in your history, which is great for improving at the game. So I'm waiting for the puzzles to be fixed, but until then I will use lichess.
Did you mean that *lichess* keeps your whole puzzle history? (Because it does -- and lets you replay any, or just the ones you missed.) On chess.com, for interesting puzzles at least, I've taken to keeping records of what I've played, which helps me confirm when I think I've seen one before. (That doesn't happen all that often, and cannot explain the huge ratings -- the only explanation for those is cheating, people who solve even the hardest problems, infallibly, in seconds. Give me a break.)
It seems like things are slowly coming around. I'm starting to see more puzzles in the 2400-2700 range--a far cry from the slew of 1600-1900 puzzles I've been seeing for the last few weeks.
Definitely seeing a change now. This has been bothering me for a while, that even at the "extra hard" setting it's not actually hard, but now there has clearly been some change. Which is great!