Why does using a set() snapshot for deduplication still allow duplicate records in my Django/Outlook integration?

I am building a synchronization script to pull emails from a specific Outlook folder into a Django database using pywin32. To optimize performance and avoid redundant database queries, I use a set() to store a snapshot of existing records for lookups. However, I am still seeing duplicate records in my database.

My Environment:

OS: Windows 10

Outlook: 2016

Tech Stack: Python 3.13, Django 5.1.4, pywin32

Database: SQLite

The Logic (Minimal Reproducible Example):

time zone

TIME_ZONE = 'UTC' # I also tried 'Asia/Ho_Chi_Minh' USE_TZ = True

model

class FloorRequest(models.Model): request_type = models.CharField(max_length=10) # 'BADGE' or 'FLOOR' email = models.EmailField() request_time = models.DateTimeField() # This is where I compare timestamps mail_entry_id = models.CharField(max_length=255, unique=True)

python

def backfill_all() -> None: items = t1_folder.Items item_list = [i for i in items] existing_records = set( FloorRequest.objects.values_list('email', 'request_time', 'request_type') ) count = 0 for item in item_list: try: sender = str(getattr(item, "SenderEmailAddress", "")).lower().strip() received_time = item.ReceivedTime.replace(tzinfo=None) # Naive datetime current_type = 'BADGE' if "badge" in item.Subject.lower() else 'FLOOR' if (sender, received_time, current_type) not in existing_records: FloorRequest.objects.create( email=sender, request_time=received_time, request_type=current_type, ) existing_records.add((sender, received_time, current_type)) count += 1 except Exception as e: continue