arcgis.gis.ContentManager.advanced_search returns duplicate results for high values of max_items

Describe the bug Content searches using arcgis.gis.ContentManager.advanced_search return additional, duplicate results in batches of 100 if max_items is set greater than 10,000 and there are fewer results than the max_items value. At max_items=10001 the result count will be incremented by 100. At max_items=10101 the result count is increased 200.

To Reproduce Steps to reproduce the behavior:

# you may need to set to an orgid you have access to
from arcgis import GIS

gis = GIS(profile='umntc_agol_admin')
print(f"Connected to {gis.properties.urlKey}.{gis.properties.customBaseUrl}",
      f"as {gis.properties.user.username} (role: {gis.properties.user.role})")

orgId = "8df8p0NlLFEShl0r"
maxItems = [300, 400, 9999, 10000, 10001, 10050,
            10100, 10101, 100000, 100001]

print("max items,result count")
for mi in maxItems:
    response = gis.content.advanced_search(
        query = f"orgid:{orgId} and type:'Feature Service'",
        max_items = mi
    )
    print(f"{mi},{len(response['results'])}")

error:

# no error presented; output follows
Connected to umntc.maps.arcgis.com as wiringa_nosso2 (role: org_admin)
max items,result count
300,300
400,373
9999,373
10000,373
10001,473
10050,473
10100,473
10101,573
100000,90373
100001,90473

Expected behavior Don’t duplicate search results.

Platform (please complete the following information):

• OS: Windows 10 64-bit
• Browser n/a
• Python API Version 1.6.2
• Python version 3.6.9
• AGOL Version 7.3 (also seen prior to the 2019-10-01 AGOL update)

[knoopum]

For anyone else interested, you can ask Esri Support to attach you to: #ENH-000144313 [Enhancement] When attempting to use the ArcGIS REST API (or the ArcGIS API for Python on top of it) in a situation where more than 10,000 items exist, the API can only access the first 10,000 items.

[achapkowski]

No, it is being looked at. It turns out not to be a Python API issue, rather a REST API issue. The max result for any query has been set to a max of 10,000 records regardless of the number of records in the query. This means there is a hard stop at 10,000. We will have logic inplace to prevent going over 10,000, and I will pass along this issue to the REST API folks.