The reason rating increases are a bad solution

The size of the pool of players determines rating variance around the mean. While you can shift the mean rating, you can't alter the spread without changing the either the size of the population or the rating algorithm itself. Adding a different amount of rating points dependent on base rating might make a few people happy for now, but it will not affect variance in the long run.

For example: Player 1 with 1800 rating points is now 2200, Player 2 with 1600 is now 1800. Skill or ELO rating difference predicts the chance of Player 1 winning against Player 2 (~75% in this case), the chance of winning then determines the rating adjustment on each outcome. But clearly the skill of Player 1 and 2 has not changed overnight and they still have the same skill level (ie a +-200 difference), but your system has arbitrarily assigned them a +-400 difference. In order for the maintain their new ratings, Player 1 would need to win ~91% of games, but since relative skill is still ~75% wins for Player 1,the algorithm will act to quickly reduce rating difference to +-200 again to reflect relative skill level as more games are played. Add more players to the pool and the effect remains the same.

The net impact will be that you shift the mean upwards by approximately the average amount of points added (e.g. 200). But since new rating differences do not represent relative skill level and the algorithm/population are unchanged, these will act to quickly bring ratings into accordance with the previous variance, making the idea of different points added as a function of rating moot. New players entering the population will also begin to shift the mean downwards unless initial rating is also shifted by this amount. If you want to adjust variance to better reflect reality, then you need to adjust the ELO algorithm such that the a constant difference in player skill results in a larger points difference.