Currently all information about current downloads may be lost if xolox crashes while it is updating the resumeinfo.txt file. (It has happened to me twice).
One quite easy way of doing it much safer is:
Rename the old resumeinfo.txt to resumeinfo.bak
Write the new resumeinfo.txt in the same way as now.
Delete resumeinfo.bak

When starting xolox:
If there is an resumeinfo.bak, use it instead of resumeinfo.txt.


Additionally it would be nice if you flushed the temporary files to disk more often than now. My experience is that you only flush the temporary files when the connection to corresponding server is lost.
Flush a file in this context => close file then open file for appending. (and update resumeinfo.txt if needed)

This has been requested before, only in a slightly different manner:

Upon starting or stopping a segment, copy resumeinfo.txt to resumeinfo.bak, then write resumeinfo.txt. If, upon startup, resumeinfo.txt is corrupt (there are fewer segment identifiers than there should be), or there are more jobs in /temp than are listed in resumeinfo.txt, copy resumeinfo.bak to txt and use that instead.

I'm not sure, but wouldn't closing the file require you to kill the connection of who you are downloading from? No program that I've used to download from the internet 'flushes' the file it's downloading in the manner you describe until it is complete or is disconnected.

Even if it could be done without dropping downloads, I don't see what this would do to help the program any. The FAT seems to be updated every time a write is done to the disk -- as I've crashed hard several times on this POS Compaq while downloading, and it hasn't found megs of lost clusters yet -- and XoloX does a rollback to eliminate corrupted datastreams.