I'm surprised you are all so calm about the paper.
I think all the privacy concerns except one were made moot by the January 2017 RingCT hard fork.
Actually most were addressed by the January 2016 fork which made mixins mandatory. The paper acknowledges this but gets it wrong in attributing it to RingCT. The fix actually happened much earlier:
The weakness studied in this section is primarily a concern for transactions made in the past, as transactions using the new RingCT transaction option are generally immune.
In fact the chart on page 6 of the paper shows the issue rapidly becoming irrelevant starting when the first MRL4 fix was deployed in January 2016.
The second part of the paper discusses temporal inferences, and while this is still being worked on the issue has also been largely mitigated.
Again, quoting the paper:
For versions 0.9.0 and prior, even up to 4 mixins, our simulation suggests that the newest input can be guessed correctly 75% of the time.
(Note that is somewhat misleading. A better wording would be that the guess is correct 75% of the time. However, you have no way to tell whether that guess is correct; you are still guessing.)
Now compare with what the paper says about the current version, as opposed one (0.9.0) from two years ago:
Under the current default behavior, i.e. 4 mixins and using the 0.10.1 sampling procedure, we estimate that the correct input ref- erence can be guessed with 45% probability (rather than the 20% ideal if all input references were equally likely).
45% is not ideal. We would like to see a 4-mixin transaction closer to the ideal of 20%. However, it is still ambiguous and mathematically untraceable (equivalent to an effective mix factor of 2.2). You can guess, and your guess will be right about 45% of the time, but you have no idea when. I.e. slightly worse than tossing a coin.
So yes, there is a lot of spin and FUD in how the paper is presented, even if its numerical results are correct (I have no reason to doubt that, but there could be errors too). It identifies no new issues at all, and the issues it identifies have already been mitigated, with existing plans to make further improvements.
