Occasional instance/memory overflows at ndb/python + manual gc.collect()'s

[Kaan Soral]

I always consider/calculate memory dynamics in my routines, since a ndb entity is maximum 1MB's, I always fetch/handle elements in batches of 20 or 40

Consider such a routine:

1) Handles 400 entities

2) Chunks 400 keys into 20's

3) In a for

 3a) fetches 20 entities async

 3b) gets the result

 3c) computes

 3d) gc.collect() ?

I think 3d should be unnecessary, yet it seems to help, since those elements are out of scope when the for re-iterates, they should be auto collected

There is also a possibility that the datastore/ndb has a leak somewhere, because If I run similar routines like this too much, I see occasional instance/memory overflows