I'd like to put to your circumstance where your statement:

"Because in every attempt to explain the functional role of feeling, feeling turns out to be functionally superfluous (except if telekinetic dualism is true, and feelings have causal ... power -- but it isn't, and they don't)."

... might be in need of rework. Feelings (P-conscious fields, in general) can be seen as causally effective in 'knowledge change'. I like to be very specific and will choose the circumstance of knowledge change to be that of the scientist who is exposed to radical novelty (there is no existing knowledge of that which is being presented to the scientist by their P-consciousness, eg a mammal with 17 legs). That is, if knowledge can be characterised as:

KNOWLEDGE(t) (1)

then the dynamic term associated with a move from ignorance to apprisal can be presented as:

dKNOWLEDGE(t)
--------------------------- (2)
  dt

Driven by the P-consciousness whiich is the scientific observcation of a novel 17 legged mammal, equation (2) drives equation (1) to 'know' the new species: DIX-SEPT-UPED. So to speak.  In the very specific case of change in the scientific knowledge, the change must be literally causally driven by the P-conscious field, because the state of the scientist's brain consistent with the dynamic revealing of the new knowledge cannot be reached (I mean this in a literal electrochemical/electrodynamical sense) unless the P-consciousness field drove it there. The effect of the phenomenal field P(t) can be detected (tested for) because of equation (2). It operates in an incredibly huge-dimensional state space where the change in possible state trajectories is driven by P-consciousness (states previously unreachable). Physically it's brain electrodynamics, but that doesn't matter at this level.

I confess a degree of frustration at the lack of attention paid to formalising knowledge dynamics (as brain electrodynamics) and the role of P-consciousness in it - especially when we have a perfectly viable way to test for scientific knowledge change. In the context of scientific behaviour, the detectuion of change in scientific knowledge supplies a prima facie case for having objectively detected P-consciousness in a human scientist (or a robot equivalent). Human science, at least, operates this way. A robot capable of the same behaviour delivers a viable claim that the robot has P-consciousness or its equivalent in a way which is independent of any internal arcvhitedtural details.......I have constructed a prototypical empirical test framework along these lines here:

Hales, C. 2009. An empirical framework for objective testing for P-consciousness in an artificial agent. The Open Artificial Intelligence Journal 3:1-15.
http://www.bentham.org/open/toaij/openaccess2.htm

I'd like to commend this line of thinking generally as bearing low hanging fruit. Equations (1) and (2) can be constructed literally into a (very very huge) set of quantum electrodynamics equations - so the formulation has a mathematical basis (as a massive 'projection operator' / 'observable' which reveals a state trajectory which traversed in the act of revealing the new knowledge). Also note that the above "knowledge" dynamics can be generalised to all belief using dynamic doxastics thus:

dBELIEF(t)
----------------   ...   (2)a
  dt

It's just that scientific belief is objectively testable (a test for acquisition of a 'law of nature'), so (2)a is not empirically useful because all they predict is a holder of a belief. NOTE: philosophical 'isms are beliefs of kind (2)a. All you need to do is act scientifically.I commend the knowledge dynamics idea to you. Note there is also a possible formulation for the 2nd derviative....

d^2KNOWLEDGE(t)
---------------- (3)
  dt^2

I leave you to ponder what (3) might be an indicator of....I have my own ideas! Q. What kind of cognitive agency results from (3) = 0?

Thus I find I cannot dismiss the causal efficacy of P-consciousness as easily as you do (not without denying my P-consciousness, and yours, and then demanding it be used for all scientific observation on pain of scientific suicide!). The self-referential application (an act of doing science directed at how we do science) seems to be the key to it. It changes the explanatory domain to one of "what perspective must I adopt in order that what exists (which is described through the P-consciousness of scientists, and is causally efficacious) might be seen to have a 1st person perspective". Which has been rather helpful ... for me, anyway.

regards,

Colin Hales