Sequences cannot have multiple headers, AFAIK. Are you sure those are not just headers with empty sequences?

You can use a combination of bioawk and sed to work for you. The sed command would be like:

echo $header | sed -re 's/^>.*gb[|]([A-Z0-9.]+)[|].*[[]([aA-zZ ]+)[]]/>\2| \1/'

HI,

You could use biopython, something like (ive done no testing)

from Bio import SeqIO

fasta_sequences = SeqIO.parse(open(file),'fasta') 
output_file="editedHeaders.fa" 
myRecords=list() 
for fasta in fasta_sequences:
            originalID=fasta.id 
            newIDList=fasta.description.split('[')
            species=newIDList[-1].replace(']','')
            newID=species + "|" + newIDList[0].split('|')[3] 
            fasta.id=newID
            fasta.description=""
            myRecords.append(fasta)

SeqIO.write(myRecords, output_file, "fasta")

I would also add a dict to save the original headers and pickle it so you can look up if needed.

Good luck