It's part of the IP and patents within NimbleOS (aka CASL) so there's no publicly facing documents that detail sweeping. However what can be said is that NimbleOS is a log structured file system which has built-in system wide garbage collection. As blocks are purged/deleted, the filesystem will re-load fragmented stripes into DRAM, perform a re-sequentialisation of said stripe into a new full RAID stripe and then restripe across the disk media. This runs all the time - it's not a scheduled process. It allows the capacity of the system to be populated up to 90% full without problematic issues such as fragmentation, read/write latency or outstanding disk queues.
If you attend one of Nimble's NTS 1-day training courses we go into it a bit deeper there as part of the file system overview.
It's a bit old now - but there's a CASL Deep Dive video on YouTube which may provide you with answers.