Build 3.5.0.73 of xlsgen makes a number of improvements in speed for working with XLSX files.

Included is :
- Faster reading of depending parts when there are many depending parts (for instance many external references in spreadsheets).
- Faster opening in Excel of files created/updated by xlsgen by writing additional metadata in the file such as spreadsheet dimensions and row spans.
- Automatic removal of useless cells, whose number may be arbitrary (and a consequence of Excel creating orphans) which lowers the file size as well as the reopening speed in Excel
- Removal of rows referencing useless cells. Again the amount being removed is arbitrary and can mount to millions.
- Better handling of out of memory scenarios (2GB is the maximum working set per process on WIN32 systems)

Build 3.5.0.72 of xlsgen makes sure that no out-of-range internal indexes remain as such in XLSX files when they are read and written back. This automatic fix allows to remove one reason why files like this may or may not be seen as corrupt next time they are open in Excel (depending on the version of Excel you are using).

Build 3.5.0.71 of xlsgen ensures that a number of Unicode characters known as surrogates (in the interval [D800-DFFF]) are encoded properly in XLSX files. These are characters that, for some reason, are expected to be of the form _xHHHH_ internally and it is an undocumented behavior. These are often used in Hebrew spreadsheets.

Build 3.5.0.70 of xlsgen ensures that the ratio in width between the sheet tabs and the horizontal scrollbar in Excel is preserved. This is a figure saved by Excel inside the file, and was ignored by xlsgen until now.