Intelligence puzzle, how to solve it?

Question

Database

Given:
Million tags
Million users
Each user can select any number of tags, creating a unique group from them and subscribe to this group

Task:
When content is created, for example, with n tags, you need to find all groups that can be combined from these n tags in an amount of 2x or more up to n respectively, and select all users who have the same groups.

What approaches, sorting algorithms and database type should be used to optimize calculations?

Answer 1

Write a solution "on the forehead", do load testing, use it to determine the bottleneck, optimize it. And so repeat a bunch of times.

Answer 2

When saving the user's settings, we hash the group of tags to which he subscribed and save the hash. When adding content with a group of tags, we also calculate the hash and look for all subscribers with the same hash.
update:
I think I got it.
Create a table of subscriptions (subscribers) of the form: user, tag
Next, make a selection from this table, something like this:

SELECT user, count(tag) as tagcounter
FROM subscribers 
WHERE tag IN (%taglist)
GROUP BY user
HAVING tagcounter>1

where %taglist is a list of article tags

Answer 3

Million tags
Each user can subscribe to any group of tags

IRL won't do that.

Million users

and this may not be, if not bots

Answer 4

There are so many excellent analyzes of such architectures on YouTube, you would know. In short, it will only be a separate subscription service. It listens to events from the system via the Event Bus and generates lists and notifications for itself. In general, queues and multiple subscribers are involved for parallel processing.