On a system with NVMe/FC, if the rport dev_loss_tmo is set to 4294967295 on all rports
(which requires that fast_io_fail_tmo be set previously, in this case to 15), if the host side
FC switch port is diabled, after 600 seconds a message similar to this appears on all
NVMe/FC controllers:

[ 959.696856] nvme nvme8: NVME-FC

: dev_loss_tmo (-1) expired while waiting for remoteport connectivity.

This is not actually true. What has actually happened is that the ctrl_loss_tmo,
which defaults to 600 seconds, has expired. (The 614 seconds appears to be due to
gross inaccuracies in the kernel timer mechanism, longer duration timers expire at
up to 2-3% beyond the time requested).