I think the problem is that in a startup, the leads are typical code slingers who are very good with programming, competent with SQL, and manage basic DBA (hand held by documentation).
Even if they are that rare startup that has heavy weights on both sides of the data/process divide, the second issue is that intelligent schema design (imo) does not fit the agile model quite that well.
Its really more of a economic issue than a technological one: it is quite possible that the poorly designed and implemented SQL - see above - solution can not scale. Obviously the problem is not SQL or RDBMs. But when you have a very smart, but young and not yet seasoned, tech team who hasn't seen it all, do you think they will blame themselves or lament that "SQL can't scale"?
