Build 3.5.0.69 of xlsgen improves the memory performance for working with XLSX files.
The reading phase is made more optimal by unzipping chunks of internal parts instead of whole parts. Working with whole parts proves not scalable for any large file such as the ones coming from the US open data government website. So this is the reason why we are not doing it anymore.
This improvement avoids "not enough on-board memory" scenarios for running such XLSX scenarios. We think any of our customers running large XLSX files will be happy of the improvement.