Lost tuple in pickle transition? (SqlAlchemy -> Redis -> Pickle - sqlalchemy

Question

Here's a tricky one.
I have a couple of VM's (debian and lubuntu), both boxes run the same code, where I pull from mysql some data and serve it as a pickled object. Everything fine in here, problem is, when pulling this info and responding it via jsonify (flask joins the game) the result is different between these two boxes... I know, not a code question... or is it? Actually I'm scratching my head for hours now, might have lost some hair in the process, can't figure out what am I doing wrong, I'd blame the distro differences (python versions is the same in both, lubuntu has a greater sqlalchemy version though) but I am not positive about it since I pulled from both responses a value by it's key (changed the query(...).all for query.first() and added a logger.debug(response.id))
Might answer some rising questions here... I "pickled" the query response for storing it in redis so I can cache that datasets instead of hitting the db constantly (didn't want to bring the topic earlier to save some confusion, the why I do this is to keep the keyedtuple as it is instead of getting back an ugly string)
# here happens the query
try:
connection = create_session(engine)
vendor_ = Table('vendor',MetaData(engine),autoload=True)
result = pickle.dumps(connection.query(vendor_).all())
connection.close()
in the view this is how I pull the info:
def all_vendor():
response = pickle.loads(controls.vendor.all_vendor(True))
return jsonify(vendors=response)
The response in the Debian box:
{
"vendors":[
{
"id": 9,
"name": "TEST",
"contact": "TEST",
"phone": "888888",
"email": "test#mail.test"
}
]
}
The response in lubuntu:
{
"vendors": [
[
9,
"TEST",
"TEST",
"888888",
"test#mail.test"
]
]
}
Weird huh? anyone has experienced this before? if so, how the * did you fix it?

Answer 1

Still unanswered. Well, it is a library thing. strace didn't help as I wanted (not and expert though) but after splitting the whole process, the retrieved by SQLAlchemy is not the same along different distro's; had to check pickle, flask and redis before getting to the assumption of the ORM being the culprit here.
No idea of the main difference in those, in any case the simplest solution was to map the whole resultset right before pushing the data to jsonify. It is not nice since it eats more resources but at least it prevents the mess.
It worked right in: Debian latest (unstable), Ubuntu 14.10.
Didn't work in: Lubuntu 14.10, Debian stable, Centos 7
That's It. Won't dig anymore into this issue.