entrez_etoolz.py

from irs import *
from urllib import urlopen

gids = ['AB019734.1', 'X86636.1', 'X86650.1', 'AB009938.1']

def rettaxids(gids):
    ''' Retrieves TaxIDs given a list of GeneIDs '''
    
    # Always tell NCBI who they're messing with
    email = 'harekrishna@gmail.com'

    addr = 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?\
email=%s&db=nucleotide&rettype=gb&retmode=text&' % email
    
    # Get the junk
    h = urlopen('%sid=%s' % (addr, ','.join(gids)))
    records = h.read()
    
    acc_to_tax = {}
    
    # Dumb and fast!
    for record in records.split('//\n')[:-1]:
        accession = record.split('\nVERSION')[1].split()[0]
        taxid = record.split('taxon:')[1].split('\"')[0]
        acc_to_tax[accession] = taxid
    return acc_to_tax
    
    
def chunks(l, n):
    return [l[i:i+n] for i in range(0, len(l), n)]    
    

for chunk in chunks(gids, 350):
    print rettaxids(chunk)

If there's an error - the bad query won't be in the returned dictionary. So you can just catch a KeyError or something.