Have you looked at:
D.1.1 Setting Up a Multi-Instance Primary with a Single-Instance Standby
Data Guard Concepts and Administration 11g Release 2 (11.2) E25608-04
If yes, what would parallel recovery give you?
I have never tried to do this setup but the main issues I would see it the second standby instance would not be available
if something happened to the first standby instance.
By using only one instance for apply it's much simplier to keep track of the redo coming from the two primary instances.
If you don't use a standby as a reader database it does not have a lot of load on it. Generally if you check the alert
log there are several minutes between the "waiting for log..." messages.
Do you have redo generation so high that one server cannot keep up? If yes I would look at setting LOG_ARCHIVE_MAX_PROCESSES higher.
Compression is another option. You can also Configure the TCP "Send" and "Receive" Buffer Sizes (RECV_BUF_SIZE / SEND_BUF_SIZE)
Check you SDU Size and increase if needed.
If you want an exact reason for you first question I think somebody at Oracle will have to answer that.