Actually the C=O group is not that stable--that's why there is so much chemistry surrounding the carbonyls in the aldehydes, ketones, and acids. Here, in this problem, the faux pas was to use the Grignard reagent with an alcohol, which are not compatible. This is why standard Grignard reactions are carried out in aprotic solvents such as diethylether and THF. Here, a secondary alkoxide anion is generated, and the subsequent acid workup restores the alcohol to its original form.