I don’t have a great or satisfying answer for you. They just tend not to, or in cases where they do, the yields are low and are accompanied by other products.

The best substrates for SN2 reactions are alkyl halides, and generally what tends to happen is either no reaction, or the Grignard reagent reduces the alkyl halide to an alkane. Deprotonation can also happen.

Part of the reason is that the structure of Grignard reagents in solution is not simple – they form clusters – and this is not good from a steric hindrance standpoint. Also, what makes Grignards add so well to the C=O group is that the magnesium can act as a Lewis acid and activate the carbonyl towards nucleophilic attack. The halogen of alkyl halides are not activated (i.e. made into a better leaving group) by magnesium.

In order to do SN2 reactions, the best choice is to use a Gilman reagent (organocuprate). Those work well, especially with primary alkyl halides. They are less basic and perform fewer side reactions.

I doubt it is a steric hindrance issue. I wonder if it is a Grignard issue. I generally have seen better results using the organolithium reagents.

Generally Grignard reagents do not react with ethers… e.g. diethyl ether and THF are commonly used as solvents.

For primary or secondary amides – deprotonation. For tertiary amides they do not generally give addition products. One exception is Weinreb amides. https://en.wikipedia.org/wiki/Weinreb_ketone_synthesis

Grignards can react with O2 to give hydroperoxides, which can then be reduced to alcohols. See https://pubs.acs.org/doi/abs/10.1021/ja01627a069

No, in that case you’re forming a C-OH bond, not a C-C bond, so you’d need to add a source of electrophilic oxygen.

How many equivalents of CH3MgBr?

Due to keto-enol tautomerism pent-2,4-dione has an acidic enol O-H that may be deprotonated faster than addition to the carbonyl can occur.