You are right @RandallWilliams, thank you for the right direction.
Root cause is our F5 load balancer health monitor's health check URLs.
As explained here, The BIG-IP system, by default, uses HTTP 0.9 when sending monitor requests, if HTTP version is not explicitly mentioned in the request.
Finally, the below health check settings worked well with F5 load balancer's pool for Workflow Manager Servers.
Send String:
GET /workflow/healthCheck HTTP/1.1\r\nHost: <FQDN>\r\nAccept: application/json\r\nConnection: Close\r\n\r\n
Receive String:
200