There are many
"phylogenetic" proposals (e.g. recent cladistic models computer-generated by Don
Ringe) with Germanic and Balto-Slavic as sister clades (the nearest relatives of
each other). Such models typically regard satemisation as an innovation that
happened somewhere up the evolutionary path (in the common ancestor of
Balto-Slavic, Indo-Iranian, etc.), only to be later cancelled in Germanic. Of
course what you get from a computer depends on what you feed into it -- here, on
your selection of "characters" to be analysed. Any bias in this selection is
likely to be reflected in the results. I don't personally regard 100%
"desatemisation" of an originally satemic langauge (combined with the
mysterious reintroduction of labiovelars to boot) as a realistically imaginable
process.
Piotr