This sounds like a problem with the ODBC driver misreporting how much memory is required for the records it intends to return. There are ways to address problems like this which I can guide you through. However first I would like you to contact support so they may work with you to collect log files, which will be necessary to determine the root cause of this problem. Please mention that Robert Morton is interested in taking a look at the logs, and I hope to get back to you soon.
In the meantime I'll try to reproduce the problem locally with the information you have provided.
Please remove these attachments from the forum -- it's best to only share those files directly with Support.
From a cursory inspection it appears that even your aggregate queries are returning nearly 10 million records, which is very suspicious. Consider examining your ODBC configuration for 1010data to determine why the aggregate queries are yielding such fine-grained results.
Last, I noticed that certain fields like 'pccoup' appear as floating-point values, but you have elected to treat them as discrete dimensions. Since floating-point fields often contain a large number of unique values, attempting to use them as a dimension can lead to very large resultsets. This may explain why your aggregate queries are failing to group results into coarse-grained buckets. If these fields are not numeric in nature (even if they are reported as such by the driver), you may need to consider creating calculated fields which cast those to INTEGER or STRING, and use only the calculated fields in your visualizations.