As Brad Larson points out in the other discussion you link, using this for Off Topic closures has become somewhat tricky:
The two close reasons that was applied on before: "Off topic" and "Not a real question" were intended to indicate things that were clearly not suitable for the site. "Too localized" didn't, because many of those questions were decent and appropriate for the site, but too narrow in scope. The off-topic variants now include questions like this, or others that may be OK questions but don't fit here. I'm not sure that downvotes should be applied to those now that they're under the "Off topic" umbrella.
Some OT reasons - on some sites - definitely do imply that a down-vote is warranted, if not several - but this is considerably less true than it was two years ago. And it may be less necessary as well...
Let's face it: this was added for Stack Overflow, which at the time was suffering greatly from an influx of low-quality questions (which hasn't changed) and a seeming unwillingness from the community to use the most obvious form of feedback (voting). A system for blocking low-quality askers had previously been implemented, but without a sufficiently-strong signal (voting) to work off of it wasn't having the desired effect.
This situation has changed fairly dramatically over the past two years. Here's a graph showing the % of closed questions down-voted to a score below 0, as well as that % when excluding the automatic down-votes from Community:
I think it's safe to say that we don't need the automatic vote to help identify problem users anymore. What's harder to say is whether or not the psychological effect of the automatic vote is still having a positive effect on the folks asking these questions... Or the folks reading them.
Still, I think we can afford to remove it at this point, if only to simplify slightly what has become a fairly complicated system. If the trend shown above reverses itself, we can always add it back...
Update: as of August 5th, Jarrod has removed this - automatic downvotes will no longer occur after the next build.
Since this change was made, question score for closed questions has remained fairly stable at around 70-75%. This seems acceptable.