Improve scalability of activity/notification fan-out for large shares and frequent updates #2660

This issue is partially related to nextcloud/notifications#2897 but with more context.

Problem

There are several scalability and UX issues in the current activity/notification model when a file is shared with large groups or updated frequently.

Scenario 1: Duplicate activity entries for group shares

If a file is shared with a team/group of 50 users, every file change creates 50 separate activity records (one per user).

This causes:

• excessive database growth,
• redundant duplicated records,
• inefficient activity storage.

A possible optimization would be to store a single activity entry for the group/share target and resolve visibility dynamically when rendering the activity stream.

However, there are some non-trivial edge cases:

• nested groups,
• users belonging to multiple groups that have access to the same file,
• preventing duplicate entries in the activity feed.

Scenario 2: Synchronous fan-out is slower than the main action

When an entity (for example, a file) is shared with many users or a large group, writing activity entries and notifications can take longer than the actual primary action (saving/updating the file) 🙈.

This becomes especially noticeable for large shares.

Would it make sense to move activity/notification fan-out into background jobs (queue-based processing), so the main user action completes faster?

Scenario 3: Excessive noise from aggressive autosave (Collabora)

With aggressive autosave settings in Collabora Online (for example every 60 seconds), each save generates:

• a new activity entry,
• a new notification for every user with access.

This creates a lot of noise and can quickly flood activity feeds and notification centers.

Possible improvements:

• rate limiting similar events,
• aggregation ("Document X was updated 15 times in the last hour"),
• debounce windows for repetitive updates.

Expected outcome

A more scalable and less noisy activity system for:

• large group shares,
• high-frequency collaborative editing,
• reduced DB pressure,
• better user experience.

Let's think/discuss together how can we improve that. I can participate in some of that improvements as well (but let's agreed on solution first)