The format between the heavy forwarder and indexer is "cooked" - which means the data after parsing, along with the metadata. All the parsing configurations need to be set on the heavy forwarder (props.conf, transforms.conf). However, some settings may need to be on the indexer or search head. While you can figure out the differences, I think it is just easier to have a duplicate of the props.conf and transforms.conf in both places - Splunk will ignore any settings it doesn't need.
If you are keeping a local index on the heavy forwarders, then it isn't really just a forwarder is it! Regardless of where you index the data, the format will be the same. BTW, if your heavy forward is set to "index and forward", it will need a Splunk license.
Yes, you can point a universal forwarder to a heavy forwarder. It works great. Just be sure to set up the receiving port on the heavy forwarder, and well as outputs.conf on the universal forwarder.
