The documentation is a little unclear. To clarify, if the vFiler owns only the qtree, not the volume then qtree level mounts must be remounted. In MOST cases, the vFiler owns the volume. A vFiler can be assigned either volumes or qtrees. A best practice is to assign volumes. When the vFiler owns the volume, both volume and qtree mounts stay mounted. When the vFiler owns the volume the vol fsid moves and all vol/qtree mounts stay intact when they are moved with the file handles to the target.
So, if you see under "vfiler status -a" that the vfiler is assigned volumes /vol/volname and not qtrees... (qtrees in volumes are automatically owned by the vfiler since below the volume assignment) it will keep mounts.
Now to the outage part... vfiler migrate does not guarantee a failover time. It runs until completion as long as it takes the final mirror and start on the target. This may or may not complete in your nfs timeout window. To get a guaranted migration within 120 seconds (using semi-sync snapmirror and the automation to fail back to source if not complete in 120 seconds), we need to integrate with OpsManager (OnCommand) and use Data Motion (won't work on 8.0, a 7.3 feature and soon an 8.1 feature).