I have some volumes that are exhibiting strange behavior. For example...
I have an 8 TB volume and it contains a single 8TB LUN. Both are thin, and nightly de-dupe enabled on the volume. The ESX datastore is VMFS 6 and has space recovery enabled. Currently I have 3.4TB free on the LUN, and would expect to have more than that free on the volume, but I don't. I have 1.37TB free on the volume. There are no snapshots.
Some pointers, do you mind taking a look at the following article, even though it has a different cDOT & ESX version but it's explaining about 'How to recover space on thinly provisioned esx lun/volume from the backend storage.
I can't think of any reason not to enable it if your use case supports it (which it sounds like it does) and the workload consuming space on the LUN can accommodate the downtime needed to offline the LUN and enable the setting.