When refreshing an extract especially with a considerably large amount of data, it eats up memory. I have had these issues with POS data. Best way is to publish the extract to Server and schedule a "Refresh Extract" so this is all done on Tableau Server, which usually has much better hardware configuration.
Hope this helps.
Have you tried hiding all unused fields and are you doing a complete refresh or an appended/incremental refresh? A full refresh might also be the issue whereas an incremental one will only update the newest records. Depends on your use-case scenario. Make sure there is enough contiguous free disk space for the largest extract in use in order to optimize extract performance. One way to achieve this is to use SSD drives as they do not become fragmented (a function of how data is retrieved from an SSD, not of where it is stored). You can also try to "Optimize" the extract as one of the options under Extracts.
Some other tips located here: Tableau Data Extracts - Tips, Tricks and Best Practices | Tableau Software
If this helps in resolving the issue, please mark this response as correct. Thanks, Don
Belated thanks. I am going to try this...I will post here my results.