I believe the routing information of the*origin* device is used - not
the destination. i.e. if you can't get*from* the Mac*to* the IBM i -
the Mac's routing information seems suspect, not the IBM i's.
The IBM i's routing information would come into play when you're trying
to get*from* the IBM i*to* the Mac. "ping"ing the Mac's I/P address
from the IBM i might be a useful datapoint... Be aware in the Mac
Firewall settings there is a "stealth mode" that, if enabled, will
prevent the Mac from responding to a ping - so you'd want that off when
you try to ping the Mac.
From the Mac, can you ping the*router* in between the Mac and the IBM i?
You're sure there isn't a firewall in front of the IBM i that
blocks/allows incoming traffic based on source I/P addresses?