Since spreading large deposits among different banks has always been the only way to keep everything FDIC insured, and since that's been a common money management practice long before CDARS existed, how did CDARS not be considered obvious and duplicating practices already in use?
Looking specifically at the '522 patent, the key feature that was viewed as new and non-obvious had to do with an automatic arrangement between a set of unaffiliated banks to reciprocally move depositors money between them and keep track of not hitting the limits taking into account that a depositor might already have accounts at some of the banks. As mentioned in my comment, this application was done under an accelerated program that requires the applicant to put themselves on the record as to what they think they have that is new and non-obvious.
In the normal case an applicant submits, the examiner rejects on some particular grounds and the applicant only responds to the rejection. With this type of accelerated examination the burden is on the applicant to do a comprehensive search (the file history has a 28+ page list of all the search strings used and what it turned up). Then the applicant has to point out what is new above and beyond what was found. From the accelerated examination support document, they think the closest references are:
Again, in normal U.S. prosecution there is no requirement for the applicant to point out what they regard as the closest references.
Their highlighting (underlined in original - bolded here) on what makes claim 1 novel is:
It seems like the basic idea is that you, the depositor, go into one bank and give them a lot of money and they, transparent to you, automatically spread it out to other banks, while the other banks are doing the same. And they figure out the interest rate they can give you by using a weighted average of the various interest rates the other banks provide. Everybody wins. Of course the two independent claims are the broadest and therefor the most susceptible to, at some point, being knocked out. However the dependent claims do go into quite a bit of detail.
The other two patents went through the system normally and cover other aspects - I have not looked at them in any detail.