my thinking is in both instances the result will be visually close or identical, but mathematically night & day. i also suspect the algorithm employed by specific encoders in enough of a wildcard to render all conjecture useless. in that case, i'd like to hear practical result-driven observations from people who've done this

if i encode a medium-quality pre-filtered source in mpeg2, decode the result(in yuv colorspace), then re-encode using the exact same encoder & settings....discounting any bit errors introduced by the idct algorithm, how mathematically identical will the result be. close? not close at all, but still visually identical or not?

if i encode a medium-quality pre-filtered source in mpeg2, decode the result(in yuv colorspace), phase shift all bottom fields 1 frame(no colorspace conversion), then re-encode using the exact same encoder & settings....discounting any bit errors introduced by the idct algorithm, will the alternating scan nature of mpeg2 encoding still preserve the mathematical signature of the individual displaced fields