It is hard to predict the progress of technology, so - NO: I won't tell you "how long before...." But I'll try to explain why CRISPR is special enough to be exciting in my experience and what technological/engineering hurdles need to be overcome in order to reach your objective.
At the moment, variations of the CRISPR-CAS system can only edit the genome of individual cells in vitro with varying efficiency. This is assuming you can culture the cells to begin with. For example, I work with human embryonic stem cells, which are particularly finicky. They won't tolerate much roughness and will even up and die on you if the growth conditions are just a bit off. This is very hard to achieve reliably as some culturing reagents (coating matrix, for example) are "undefined" products with variations in composition from batch to batch.
To go to a chop shop and treat your issue at the genetic level requires an in vivo way to introduce a CRISPR-enabled vector into your cells. This is not easy to do with today's technology, but it may not necessarily be a deal breaker. In the example you gave, a food allergy can probably be addressed by treating only the GI tract and the immune system that comes into contact with the offending allergen. As such, there is no need to target every living cell in your body in this case. However, if you are treating an illness involving a more fundamental life process, that is not the case. For example, a mitochondrial disease where basic cellular metabolism is defective would probably be best tackled when an individual is still a developing embryo or at least very, very young. Otherwise, tissues and organs that are not convenient to access will still retain the genetic defect and present problems for the host organism.
Another question is where in the genome you want to edit. So far, one of our experiments involving the targeted insertion (non-CRISPR method) of a construct into our hESCs have been a bust. Our best guess is that the intended site of transfection (sub-telemeric regions of chromosomes) is critical for cell survival and too much fiddling in the area is fatal. CRISPR-CAS was a compelling solution for us because of how ideally targeted it is supposed to be. We are not aware of anyone else who've used CRISPR with hESCs in the way that we are doing, but what has been reported so far with other experiments using notoriously difficult subjects has been encouraging. So far, the experiment shows clear evidence of true integration into the genome as opposed to a transient transfection. In about a week, a Southern blot verification will tell us if the integration was random or indeed targeted.
As rosy as I can paint a picture about what is possible, however, strong caution follows the introduction of any new technology. Anonymous Coward may be an asshole, but (s)he isn't wrong for being a cynic about the commercial deployment of this as a consumer product. Considering how complex human biology is, the chance of an unintended edit with unanticipated consequences is more than likely. Many genes are linked in very convoluted ways. Even with the human genome project having ostensibly mapped everything, we are still looking at just the tip of the iceberg. Having a complete manuscript, is very different from understanding all the nuances of the story. To get back to the spirit of your question, I would imagine that the scenario probably is more similar to dental service, where you go back periodically to check on the integrity of any major service, with tweaks along the way as necessary.