Details. So the consecutive loss 'penalty' seems to work the following way. If player A has a win streak against player B, and player A wins, the ELO changes are reduced for both of them by the same percentage. But if player A has a win streak against player B, and player B wins their next game, they both get the full ELO change.
On one hand, the ELO leak has stopped, because the rating that was denied from player A goes to player B instead. So at least not everyone's rating will go down in the long run. However, there are still big issues with this approach. If player A has a 6+ game win streak against player B, then player A has nothing to win and everything to lose against player B, so player A is disincentivized from playing player B. Also, if there is still a 30-day cutoff on the win streak count (I can't tell for sure yet, but I assume this is the case), then stronger players (who typically have win streaks against others) are still incentivized to play less until the win streaks clear.
Hoever, the long term consequence of the current version is that strong players who keep playing through win streaks will be punished (their ELO lowered), while weaker players will have their ELO increased, so overall, ELO will be a worse indicator of skill. (As a side note, the winning chances displayed in the 'ELO calculation detail' box will also remain incorrect). Also, strong players who play fewer games will be less affected by the win streak penalty, so their rating will be higher than the rating of players of the same skill that play more games. So you still managed to create an unfair playing field where ELO reflects not only skill, but also how frequently a player plays.
A possible remedy is to apply the % penalty regardless of the result of the current game. This way, if one of the players has a win streak, the stakes will be lower for everyone. This would be effectively equivalent to lowering the K factor. If we keep pretending that this change is to hinder cheating, you would need to change the % penalties to be bounded away from 100%, otherwise a cheater can just transfer ELO to one account until 100% penalty is reached, then lose one game with no ELO exchange, then keep repeating. If the penalty is bounded away from 100%, this is less of a problem as the one loss would erase a lot of the ELO gain.
An even better option is to set a K factor decrease for frequent opponents regardless of any win streaks (or who wins in general). The K factor should not decrease to 0 as that would make games friendly; instead, gradually decreasing to, let's say, 20% of the original K factor would work well. With this approach, you can also keep the 30-day cutoff for the frequent opponent game count. The long term effect of this solution would be that players converge slower to their true ELO, but everyone would have the correct ELO eventually, while it would still be hard to gain a lot of ELO against a single opponent.
So, to summarize, even though the intentions were good this time, you still managed to mess up again, despite the fact that the specific pitfall above was pointed out in this very thread within 3 hours (!) of the last official update. I am at a loss for words.
Either way, another update is in order. The lowered K factor would be a reasonable compromise: it still makes it hard to gain a lot of ELO against a single opponent with minimal side effects (that I think most players here could live with). Honestly, at this point I would also consider just scrapping the whole win streak penalty system and look for other ways to fight cheating. Just return to the regular ELO, because whoever makes the decisions is clearly incompetent to carry out a proper analysis of the long-term effects. Don't tinker with what you don't understand.
At least make it display what the base value was.