PowerPath fails to claim devices during boot

blade servers with 650 FLB CNA's have problems claiming devices for powerpath.
powerpath fails to claim the devices and then it defaults to NMP.
This happens at random at boot time.   Sometimes it works, and sometimes it doesn't.
When claiming fails it will sometimes only fail on a few luns, but other times it will be all luns.

these are the errors seen in the logs:
2016-02-07T16:46:14.386Z cpu10:33310)ALERT: PowerPath:MpxRecognize failed. Path vmhba0:C0:T1:L1 not claimed
2016-02-07T16:46:14.386Z cpu10:33310)ALERT: PowerPath:Could not claim path vmhba0:C0:T1:L1. Status : Failure
2016-02-07T16:46:14.386Z cpu10:33310)ScsiPath: 5562: Plugin 'PowerPath' rejected path 'vmhba0:C0:T1:L1'
2016-02-07T16:46:14.386Z cpu10:33310)ScsiClaimrule: 1388: Plugin PowerPath specified by claimrule 250 was not able to claim path vmhba0:C0:T1:L1: Not supported
2016-02-07T16:46:14.387Z cpu10:33310)ALERT: PowerPath:MpxRecognize failed. Path vmhba1:C0:T3:L6 not claimed
2016-02-07T16:46:14.387Z cpu10:33310)ALERT: PowerPath:Could not claim path vmhba1:C0:T3:L6. Status : Failure
2016-02-07T16:46:14.387Z cpu10:33310)ScsiPath: 5562: Plugin 'PowerPath' rejected path 'vmhba1:C0:T3:L6'
2016-02-07T16:46:14.387Z cpu10:33310)ScsiClaimrule: 1388: Plugin PowerPath specified by claimrule 340 was not able to claim path vmhba1:C0:T3:L6: Not supported
2016-02-07T16:46:14.387Z cpu10:33310)ScsiPath: 5549: Plugin 'NMP' claimed path 'vmhba1:C0:T3:L6'
2016-02-07T16:46:14.388Z cpu10:33310)ScsiPath: 5549: Plugin 'NMP' claimed path 'vmhba0:C0:T3:L1'
2016-02-07T16:46:14.388Z cpu10:33310)ScsiClaimrule: 1165: The current claimrules indicate that path vmhba0:C0:T3:L2 should be claimed by plugin PowerPath.
2016-02-07T16:46:14.388Z cpu10:33310)ScsiClaimrule: 1169: Path vmhba1:C0:T3:L2 which appears to refer to the same physical media as path vmhba0:C0:T3:L2 is already claimed by plugin NMP.

Manually unclaiming the lun and reclaiming it will work normally leaving the lun under powerpath control.  Claiming process only fails when it occurs as part of the boot sequence.

When ESXi offers a path, PowerPath sends Inquiry command (0x12) as part of path claiming process. From debug logs, PowerPath is receiving Inquiry failure with Host error - VMK_SCSI_HOST_NO_CONNECT (H: 0x1) for all the paths not claimed by PowerPath.

2016-05-22T15:25:41.502Z cpu18:33310)PowerPath:PowerPlatformScsiIoErrorIsRetryable: cmd=0x12 Failed H: 0x1 S: 0x0 P: 0x0 Path=vmhba1:C0:T2:L500
2016-05-22T15:25:41.502Z cpu18:33310)PowerPath:INQ failed and NOT RETRIED on device: vmhba1:C0:T2:L500
2016-05-22T15:25:41.502Z cpu18:33310)PowerPath:Inquiry Pirp Failed: vmhba1:C0:T2:L500

In general, HOST_NO_CONNECT status is returned if the connection is lost to the LUN. This can occur if the LUN is no longer visible to the host from the array side or if the physical connection to the array has been removed. 

This is a problem with the CNA card during boot time.  This ends up sending the NO_CONNECT message to the OS causing it to fail to claim. 
This is not a PowerPath issue. Inquiry has to succeed for PowerPath to claim a path.

Upgrade CNA drivers or patch ESXi.

ESXi 6.0 U2 fixes this problem on HP ProLiant BL460c Gen9 blade servers.

Environment where this bug has been seen before:

VMware ESXi 6.0.0 build-3380124
HP ProLiant BL460c Gen9
Emulex lpfc OneConnect OCe14000, FCoE Initiator : 650FLB CNA 
PowerPath version 6.0