Someone finally re-posted a link to the exhaustive capacitor test I had seen and lost:
http://uk.geocities.com/cyrilb2@btinternet.com/downloads_4.html
P. 13 of this document tests in the reduction in distortion with large film bypass caps--1uF and 10uF on a 100uF electro:
http://uk.geocities.com/cyrilb2@btinternet.com/download/capsound6.pdf
The measured reduction in distortion was only 1 and 2dB. It may be worth testing your selected film and electro caps for distortion before committing space for the bypass. Note that document #5 shows that bipolar electros exhibit lower distortion than either bypassed or unbypassed polar electros.
In spite of that, I haven't moved to nonpolars even though it would make my design life easier. Some comments out there about the reverse polarity only intended to be transient rather than constant, and the resulting effect on capacitor life . . . never seen it quantified, but that combined with the difficulty in sourcing them keeps me stuck with high-quality polars and hoping for the best