Yes this could be the case if the ideal feedback point is for instance 40dB. However 37dB is pretty steep (1/75th of the open loop level) so I am thinking that perhaps I can still get a good sound at say 30dB - and of course more level.
So it's a trade off between sound quality and level. When you first add feedback (for whatever reason - to lower the impedance in my case) you can add a whole bunch of harmonics too (depending greatly upon the tube type, operating point etc etc), which then take a further amount of feedback to squash again (depending on tube type etc). Ironically the amount of feedback that is optimal is thus dependent entirely on the linearity of the open loop stage. A perfectly linear stage will take zero to lots of feedback with no distortion issues (except possible frequency dependent phase distortion).
So more feedback = lower impedance + varying linearity (worse then better)+ rising phase distortion + falling gain and noise. Bandwidth is also affected but feedback over two tubes means that open loop bandwidth was never an issue anyway here.
As for listening results the sound test will be influenced by the impedance, phase distortion and linearity - which one is more noticeable is the one to tune, and probably also depends upon the musical content too.