Does tcp wait recycle function on disabled pool nodes?

By default F5 makes load balancing decision for the very 1st HTTP request. Subsequent HTTP requests are sent to the same pool member that received the 1st HTTP request. This could cause TCP connection to persist on a pool member even when it is disabled.

I would recommend using OneConnect with /32 netmask along with "Re-select" within the pool.

This will force the F5 to make load balancing decision every time an HTTP request comes in and this in combination with "Re-select" will cause the connection to be drained from failed pool member to new active pool member.

I am assuming you are using standard VS for HTTP traffic.