Have you ever seen this error host alarm on your vSphere client console?

Yeah, you have an issue, but the host is still working, so ... first of all ... NO PANIC!

The message says that the OS (esxi) boot storage is not responding. This means that configuration changes will not be commit to the host and they will not be available on next boot.
In other terms, if you simply don't change anything on the vSphere configuration, you will not have further issue on next reboot!

Now, you should move the active VMs to other working host(s) before to analyze/repair the trouble.

Once you moved the guests to other hosts, open a ssh connection to the host to check OS in order to read debug logs and informations.
For instance, using "dmesg" you will see that the trouble is related to USB device:

Again, you can retrieve useful informations from "Integrated Management Log" on the iLO server console.

In my case I found a trouble related to "HP Dual 8GB MicroSD EM USB Kit" (P/N=741281-001):

This dongle is a very nice solution for esxi system because the OS is very small and most of space of a large traditional hard drive would be simply wasted.
You can save money and have data redundancy through a mirrored RAID 1 configuration on a small dual ­microSD card module.
Just connect the module to the internal USB port and install ESXi as usual!

Unluckily, in my case, I've got a trouble...

I've open a support case to the vendor (HPE) but, after a long queue of email, phone calls, and tons of firmware and software upgrades that didn't solve the issue, I give up to find a resolution of the problem :(
The remaining solution is to perform the following workaround steps:

1. Move guests to other hosts.
2. Power off the failed host.
3. Restart host.
4. !!! Host lost sync with the cluster. !!!
5. Remove host from cluster.
6. Connect to single host.
7. Remove zombi guests from local inventory (guests already moved to good cluster host, but the vmotion notification has not been saved to the configuration of affected host due to the USB failure).
8. Re-add the host to cluster.

Finally, few weeks ago, I found the following HPE bulletin with the acknowledge of the issue:

http://h20564.www2.hpe.com/hpsc/doc/public/display?docId=emr_na­c05369827

SUPPORT COMMUNICATION ­CUSTOMER ADVISORY Document ID: c05369827 Version: 1 Advisory: (Revision) HPE Flash Media Devices ­ When Utilizing an HP Dual 8GB microSD Enterprise Midline USB Kit On An HPE ProLiant Server Running VMware ESXi, A Memory Corruption And All Paths Down Situation May Be Experienced

The affected device have P/N=741281-001 or P/N=741281-002 and they should be replaced with P/N=741281-003 (provided at no charge).

The article says also that it's not necessary to reinstall the Operating System when replacing the Dual 8GB microSD Enterprise Midline USB Kit. It is only necessary to move the existing microSD Cards to the new dongle.

Honestly, I preferred to replace also the microSD ... see next article: Replace defective cards on Dual 8GB MicroSD with RAID1