python-flask model incorrect for multi-level $ref

Overview

Testing using v3.0.68, when a schema references another schema which is itself a reference, the resulting python-flask model does not contain all of the expected items.

Example

The following OAD is used to demonstrate the problem:

openapi: 3.0.4
info:
  title: Demo of ref bug
  description: |-
    This demonstrates the multi-ref bug
  version: 1.0.0
  
paths:
  /blobs:
    get:
      summary: Return a list of blobs
      description: Return a list of blob (or sub-class) objects
      operationId: getBlobs
      responses:
        '200':
          description: Successful operation
          content:
            application/json:
              schema:
                type: "array"
                items:
                  anyOf:
                    - $ref: '#/components/schemas/BlobA'
                    - $ref: '#/components/schemas/BlobB'
                    - $ref: '#/components/schemas/BlobC'
                    - $ref: '#/components/schemas/BlobD'
components:
  schemas:
    BlobA:
      description: A class blob
      type: object
      required: [a_item]
      properties:
        a_item:
          description: Item at A class level
          type: string
    BlobB:
      description: B class blob
      allOf:
        - $ref: "#/components/schemas/BlobA"
        - type: object
          properties:
            b_item:
              description: Item at B class level
              type: string
    BlobC:
      description: C class blob
      allOf:
        - $ref: "#/components/schemas/BlobB"
        - type: object
          properties:
            c_item:
              description: Item at C class level
              type: string
    BlobD:
      description: D class blob
      allOf:
        - $ref: "#/components/schemas/BlobC"
        - type: object
          properties:
            d_item:
              description: Item at D class level
              type: string

Expectation

It is expected that BlobA contains only A class items, BlobB contains A and B class items, BlobC contains A, B, and C class items, and BlobD contains, A, B, C, and D class items.

This expectation works in SwaggerEditor, where the schema matches the above.

Result

The resulting python-flask server stub only seems to contain one parent class worth of items.

For instance BlobD model is as follows:

class BlobD(Model):
    """NOTE: This class is auto generated by the swagger code generator program.

    Do not edit the class manually.
    """
    def __init__(self, d_item: str=None, c_item: str=None):  # noqa: E501
        """BlobD - a model defined in Swagger

        :param d_item: The d_item of this BlobD.  # noqa: E501
        :type d_item: str
        :param c_item: The c_item of this BlobD.  # noqa: E501
        :type c_item: str
        """
        self.swagger_types = {
            'd_item': str,
            'c_item': str
        }

        self.attribute_map = {
            'd_item': 'd_item',
            'c_item': 'c_item'
        }
        self._d_item = d_item
        self._c_item = c_item

Reproduce the issue with the supplied OpenAPI document and inspect the generated python-flask BlobB, BlobC, and BlobD models. Compare inherited properties with the stated expectation; done means each model contains the fields from all referenced parent levels. The issue names no source file or test to begin with.