I'm having trouble understanding why dissolving metal reductions, like (for example) sodium with alkenes in liquid ammonia, sodium with aldehydes/ketones in ethanol, or the Bouveault–Blanc reduction, reduce the substrate instead of the solvent. If you put sodium in ethanol (or any protic solvent), the ethanol is reduced and hydrogen gas is released. I can't see how the presence of a ketone (for example) will create a more favorable reaction.
For once, it doesn't seem thermodynamically favorable - reduction of the ethanol will release hydrogen gas, largely increasing the entropy, while reduction of the ketone produces no gas. Even if it is thermodynamically favorable, however, reduction of the ethanol seems much more favorable kinetically, since there are much more solvent molecules than substrate molecules. What is going on here?