If you have a gearman client like the one below and you want to submit many jobs, then you run into a memory leak.
client = gearman.GearmanClient(server)
for job in many_many_jobs:
req = client.submit_job(task, data) ...
The problem is in the python module client_handler.py, where every job is stored in a dictionary but never deleted.
Working on a patch that should help alleviate (but not completely) solve your problem.
Closed with python-gearman 2.0.2
Thanks for the bug report!