While we can certainly agree that the rules are quite arbitrary, I don't think that unplugging the roomba is the solution here. Here are some things to think about:
You say that "controversial" votes cause it to be scored negatively, and it remains unanswered even though it is answerable, just because it is "misguided". If that's the case, I would suggest taking this as an opportunity for the community to introspect and analyze its voting and answering patterns before blaming it on the roomba and asking for a system wide change.
Specifically, these are issues that you should be discussing on your meta (I don't know if you already have, since I'm not on physics.se)
- Is the community being too harsh on questions that fall into areas that they are inexperienced/disinterested in? Then perhaps that attitude could be changed...
- Is the community choosing to ignore questions that it deems not worthy? If so, then it shouldn't complain that the system agrees with it that the questions are not worthy.
In the end, while it's easy to blame the system, if you want a long term solution to the problem, some introspection is needed. Besides, now that OPs can see their own deleted questions, there's nothing stopping them from trying to ask a better version of the same question, incorporating the suggestions in the comments (if any, and if they haven't done so already).
One partial solution to this, which would help the really small sites (not physics.se) would be to add a "has at least 100 views" criterion to the deletion task. If 100 people viewed it and didn't find it worthy enough to correct the vote imbalance, then it probably ought to be deleted.