Schedule a full backup of the mailbox stores before you defrag them.
Make sure you have free space on the drive where the defrag is occurring of at least 110% of the mailbox store size.
Plan on the process being able to defrag at a rate of about 4 to 5GB per hour. This is Microsoft's guideline. The process will often run faster than this but plan on this number for your maintenance window.
Schedule a full backup of the mailbox stores after the defrag is completed. The defrag process actually replaces the original file with the defragged file. If you don't perform a full backup afterward it may throw off any incremental or differential backups that occur, depending on your backup software.
Your users may not be fully on board with the extended downtime, even if management is. You may get some requests for progress status from unusual quarters.
It'll make a full copy of each datastore file as it goes, so make sure you have enough disk-space.
If you have multiple datastores, you can run the defrags in parallel to save time. Copy the datastore file to somewhere else and run the esutil task on the other server. Copy back when done. When we landed some database corruption, this method meant we were only down for 18 hours not 36.