Summary

• systemtags activity handling for MapperEvent::EVENT_ASSIGN / EVENT_UNASSIGN currently resolves recipients by iterating all mounts returned by getMountsForFileId() and repeatedly calling getPathsForAccessList() for the same file event.
• On instances with many mounts, this creates significant overhead in Listener::mapperEvent() because access-list expansion is recomputed many times.
• This change resolves a canonical file node (prefer owner-home node when available), calls getPathsForAccessList() once, and reuses that user-path map for activity publishing.

Problem

• mapperEvent() did per-mount:
  • resolve owner folder
  • resolve node by file id
  • call getPathsForAccessList(node)
  • merge users map
• getPathsForAccessList() already performs access-list expansion for the file node, so per-mount repetition is redundant and scales poorly with mount count.

Solution

• Keep mount iteration only to find a resolvable node.
• Normalize to the file owner’s home-node when available ($node->getOwner() + owner getById()), to use canonical share/access context.
• Call getPathsForAccessList() once, assign $users = $al['users'], and stop iterating.

Why this is safe

• Event gating is unchanged (files object type, non-empty tags, assign/unassign events, activity app enabled).
• Activity publishing behavior is unchanged (same subject building, visibility/admin checks, and updateLastUsedTags behavior).
• If owner-home re-resolution does not find a node, behavior falls back to the already resolved node.

Performance impact (measured example)

• Before: user-resolution path ~34s (single event on a high-mount file).
• After: user-resolution path ~8–12ms in the same scenario.
• End-to-end mapper listener time dropped from tens of seconds to tens of milliseconds.

Test plan

• Assign a tag to a file and verify activity entries for expected recipients.
• Unassign a tag and verify corresponding activity entries.
• Verify invisible-tag behavior remains admin-only for non-visible tags.
• Verify behavior on shared files (including non-owner context).
• Verify no activity is emitted when activity app is disabled.
• Compare runtime before/after on a file with many mounts.

Notes for reviewers

• Optimization is focused on removing repeated access-list expansion in the hot path.
• Added inline comments explaining:
  • why owner-home node normalization is done,
  • why a single getPathsForAccessList() call is sufficient.