Rationale for team_memberships id-file fan-out (team_id -> team_id:user_id):

• --id-file provides a team id, but team memberships are naturally a set of rows keyed by (team_id, user_id).
• The source API is team-scoped (/api/v2/team/{team_id}/memberships), so we must fetch members for that team and expand to per-membership records.
• State/mapping for this resource already uses composite keys (team_id:user_id) via _team_membership_key, so fan-out preserves existing identity semantics and diff behavior.
• The implementation does delete-before-write for that team prefix to remove stale memberships for users no longer in the team.
• Empty team case intentionally raises SkipResource so we don't persist an invalid bare team_id row in state.

Codex Findings
MAJOR - datadog_sync/model/team_memberships.py#L102: the hand-rolled targeted pagination only stops when len(members_of_team) < page_size, ignoring meta.pagination.total. Reproducer: team has exactly 100 memberships, page 0 returns 100 members with total=100, then the unnecessary page 1 request hits 429/5xx. The code classifies the whole team as transient and drops the 100 complete members. Fix: stop when the fetched count reaches meta.pagination.total or reuse the existing remaining_func; add a regression test that page 1 is not called when total == page_size.
MINOR - datadog_sync/model/team_memberships.py#L150-L151: a valid empty team is reported as skipped/errored even though the method contract says a team ID fans out to “zero or more” memberships. Reproducer: /memberships returns {"data": []} for an existing team; the import emits a skipped outcome for a successful no-op. Fix: return success with an empty list, or emit a distinct success/noop outcome if callers need visibility.
NIT - datadog_sync/utils/configuration.py: the _ID_FILE_SUPPORTED_TYPES docstring still says “Monitors only in v1” after adding authn_mappings and team_memberships. Update that and the stale test wording around “non-monitors” to avoid future reviewer confusion.