Hmmm. Would you be happier if I changed my last line to read "... we need to discard the whole Si concept as inappropriate to our imperfectly-observed universe."?
I don't think so. Solomonoff induction applies to streams. The most common application is to streams of sense data. There is no pretense of somehow observing the whole of the universe in the first place.
You're about to flip a quantum coin a million times (these days you can even do it on the internet). What's your estimate of the K-complexity of the resulting string, conditional on everything else you've observed in your life so far? The Born rule, combined with the usual counting argument, implies you should say "about 1 million". The universal prior implies you should say "substantially less than 1 million". Which will it be?
EDIT: Wei Dai's comment explains why this post is wrong.