If an error occurs while a particular initiator is accessing
    a SCSI device, that error is latched for all other initiators
    and is not unlatched and reported to the other initiators until
    the next time they access the device. Therefore, if the other
    initiators do not access the device in a timely manner, the
    reporting of the error can be greatly delayed, which can cause
    confusion.

    The purpose of SCSI_ERROR_POLL is to cause OpenVMS to send a
    SCSI Test Unit Ready command every hour to each SCSI disk, in
    an attempt to force latched errors to become unlatched and to
    be reported immediately. SCSI_ERROR_POLL has a default value of
    1. It can, however, be set to 0 by the user in order to stop the
    error polling activity.

    The parameter affects SCSI disks connected by Fibre Channel as
    well as parallel SCSI. If the disk has multiple paths, then the
    error polling is performed on all non-served paths to the disk.
    Tapes and other non-disk devices are not subject to this error
    polling, regardless of the parameter setting.

    SCSI_ERROR_POLL is a DYNAMIC parameter.

    SCSI_NOAUTO is a DYNAMIC parameter.