A about March or earlier, the latest rules for BizTalk Health Monitor started to show the following important warning.
I had already been advising clients to run SFTP receive locations on a separate host instance due to issues where it would silently stop receiving requiring a host instance restart to resolve, possibly related to FIX: SFTP adapter receive locations stop working intermittently during higher load in BizTalk Server
The rule itself also talks about MQSC and DB, but as I didn’t have any receive locations for those it was only warning about SFTP.
I’ve seen that warning even on a BizTalk 2010 environment that has a Custom SFTP adapter installed, possibly because it was also called simply SFTP. However it does not warn about third party SFTP adapters such as the nSoftware one which has a different name.
However, even after changing the Max worker threads to 500 and restarting the Host Instance and re-running BHM, the warning did not go away. Even moving the receive locations running on the nSoftware onto another host instance, and un-configuring the adapter on that host instance didn’t make any difference.
The BizTalk adapter’s default configuration is to poll every 5 seconds. This is rather excessive and should be reduced.
Recommendations for SFTP Receive Locations
- Have a dedicated host for receive locations using SFTP
- Increase the Maximum Worker threads setting to 500 on that host instance
- Check how frequently you poll (the default is 5 seconds), increase the delay if that is not needed.
- Put a schedule on to only poll during the periods you need.
- Disable message body tracking if it is not needed.
BizTalk musings 
We had this issue on 2013R2 & 2016 & I created a dedicated host and increased the polling time like you. I still had issues. For one external SFTP Server I had to stop & start it every hour but for others I did not have to do this. I think this was a special case. But all my these issues went away with the latest 2016 CU. Furthermore I don’t see the warning you see in the BHM.