Alkali metal, main group, and transition metal (TM) alkoxides all have significantly different chemistry. Early- and Late-TM alkoxides also have different behaviour, so we should split this family into at least four categories. All of them are basic and will decompose via hydrolysis if dissolved in water; however, this type of chemistry is not usually carried out in presence of water.
In TM chemistry, alkoxide or phenoxide are mostly used as ancillary ligands, meaning they do not usually participate in the chemistry. Although they are liable to be hydrolyzed under protic conditions, alkoxides are actually known for being especially stable TM ligands because they do not undergo β-hydride elimination (BHE) (1), which is one of the most common decomposition pathways for TM alkyls. For instance; you often cannot make a stable t-butyl TM complex because the β-hydride will be transferred to the transition metal, liberating 2-methylpropene unless the metal is totally coordinatively saturated. An important example of a BHE-stable TM t-butoxide is the following Mo(IV) Schrock catalyst, which is stable:
It should be noted, though, that the Schrock catalyst doesn't actually have any β-hydrides. Later, I'll show an example of a TM-alkoxide that does.
An important feature of alkoxides as ligands for TM chemistry is that they are electronically malleable. Since an alkoxide has 5 electrons it can behave either as an σ-donating X ligand or as a σ- and π-donating LX ligand (and formally as L2X). Presumably, this electronic malleability has something to do with their stability to BHE. An alkoxide's π-donor character might stop the metal center from liberating a coordination site to accept the β-hydride. For more information on BHE-stability, see (2), as cited by Schrock in (1).
Since early-TMs are more electropositive, the bonding character of their alkoxides is more ionic, although not to the same degree as that of alkali metal alkoxides. Alkoxides are very good bridging ligands, and if they are not sufficiently bulky, alkoxide complexes often polymerize. With a sufficiently bulky monomeric early-TM alkoxide, the main decomposition pathway is formation of the oxo complex. For example; the decomposition of a metal diolate shown in the next image was observed by Gable et al. (3) Their Hartree-Frock calculations implicated an empty metal MO at rhenium in the M=O bond formation, as shown in the MO picture in the next image.
Late-TM alkoxides are expected to have weaker M-O bonds; and not as many have been isolated, but a new study (4) has examined these M-X (where X = O, N) bonds.
A really nice example illustrating a few important TM-alkoxide concepts is tantalum(V) ethoxide, Ta2OEt10. See wikipedia: https://en.wikipedia.org/wiki/Tantalum(V)_ethoxide
Ta2OEt10 is a stable TM-alkoxide with true β-hydrides, bridging oxygens, and it's readily hydrolyzed when exposed to protons. It can be synthesized via salt metathesis with sodium ethoxide, a reaction that has equilibirium lying in the opposite direction of the alkali metal alkoxide formation you proposed.
In MPV reduction (and OPP oxidation), Al(O-i-Pr)3 is usually used as the catalyst. Aluminium is a metal, but it is a main group metal, and its chemistry is somewhere between that of boron, a semimetal, and that of early transition metals, except, of course, that it does not have a partially filled d shell. Al(O-i-Pr)3 is an electron deficient Lewis acid, comparable in chemistry to boronate esters, but much more reactive. The proposed mechanism for MPV/OPP [R]/[O] goes something like this:
It is a true catalytic reaction, so the Al(O-i-Pr)3 is indeed regenerated, but I don't know if it's a routine part of the procedure to recover it. Al(OR)3's are not, however, soluble in IPA, but alcohols and ketones usually are, so you could probably precipitate Al(O-i-Pr)3 in IPA providing you have a way to seperate your ketone and your secondary alcohol afterwards.
Interestingly, Ti(O-i-Pr)4, is analogous to Al(OR)3, but its reactivity totally different, since titanium is a true TM and not just a Lewis acid main group metal like aluminium. Ti(O-i-Pr)4 is the catalyst for sharpless epoxidation, so this might be a good reaction to look at as an example of early-TM alkoxide reactivity.
To answer your questions in list form:
- Pretty much all alkoxides are susceptible to acidic attacks, but, being fairly strong bases themselves, they are probably stable to weaker bases
- Although they will react with acids, they will not necessarily react with electrophiles unless they are stronger electrophiles than the metal center; however, early-TMs are fairly strong electrophiles. I don't think people use TM alkoxides for ether synthesis though, and you always have to keep in mind d orbitals and the unique rules of TM chemistry when you're discussing TM complexes.
- As mentioned above, and in my comment; "You can't have MX + M'(OR) -> M(OR) + M'X, where M=alkali metal and M'=TM, because the equilibrium would lie in the opposite direction, since the highly nucleophilic alkali metal alkoxide M(OR) will attack the electrophilic transition metal M'X (especially if it's an early-TM). As Schrock says in (1), "the relatively high electrophilicity of an early transition metal makes dissociation of alkoxide ion, at least in a neutral species, unlikely".
- If you can find a suitable means of seperation like a solvent that will precipiate your metal alkoxide, you should be able to seperate it, although some catalysts will be deactivated after a certain number of cycles.
(1) R. R. Schrock, Polyhedron, Vol. 14., No. 22, pp. 3177-3195, 1995
(2) D. C. Bradley, R. C. Mehrotra and D. P. Gaur, Metal Alkoxides. Academic Press, New York, 1978
(3) K. P. Gable, J. J. J. Juliette. J. Am. Chem. soc, 1996, 118, 2625-2633
(4) J. R. Fulton, A. W. Holland, D. J. Fox, R. G. Bergman. Acc. Chem. Res., 2002, Jan; 35(1): 44-56