Disk write failure, NTFS Cluster size fragmentation

Customer (see linked CBSE) hit an issue where disk writes starting failing, memcache reported "WINAPI error = 665" this was narrowed down to a limitation in the NTFS filesystem. Their vBucket files had too many fragments, see:
Microsoft KB Article, Article ID: 967351, A heavily fragmented file in an NTFS volume may not grow beyond a certain size http://support2.microsoft.com/default.aspx?scid=kb;EN-US;967351

If ep-engine sees a disk write error, it would have been possible to continue operating by creating a new vBucket file instead of appending the existing one. Would such an automatic recovery action be possible to implement ?

Microsoft recommends that NTFS filesystems serving Exchange or MS SQL Server data are formatted with 64-KB allocation units instead of the default 4k to lower fragmentation. Should we make the same recommendation for Data and Index directories in our documentation ?