When is work for a client "connected to" work for a different client?

Question

In a consulting or contracting business, when is ownership of intellectual property work done for client A separated from work done for Clients B & C?

The contract agreements are usually crisp and clear about defining "Work" to included code, models and so on, and "IP" is also well defined to include patents and the client's future right to patent.

When it comes to the scope of ownership there seems to be more reliance on legal or dictionary interpretation.

The wording is something along the lines of "Client owns IP for work that is related, directly or indirectly to the client's current and foreseeable future business."

Alternatives are "IP rights arising from work performed for the client" and "IP rights in connection with work performed for the client.

Example: Client A of the contractor needs some R&D work done in the area of machine learning for scanning customer meal trays and automatic payment processing. Concurrently Client B needs work done in the area of machine learning, for verifying food packaging as part of factory automation.

Perhaps there is no "connection"/"relationship" because one is front of store and the other is factory, two different problems and each with their own idiosyncrasies. But are they "connected" because both are machine learning problems?

The question How to manage intellectual property right for different clients from a consultancy has a good answer by Ron Beyer.

If the client supplies an algorithm and it is coded in software by the contractor, the client owns the algorithm. If the algorithm is developed by the contractor for the client, the contractor owns it, but the client owns the software.

This perspective was clear to me, and contracts may draw the line differently so that the developed algorithm is also owned by the client.

The answer draws another distinction too: IP ownership is separated if knowledge is gained independently. This likely comes down to how specific the gained knowledge is.

This leads me to believe that "connected to", "arising from" and "related to" do not refer to the technology itself, e.g. machine learning, but whether they are related because of how the work output for client A was performed/shared/reused by the contractor for client B.

Another good answer is provided by Ohwilleke in What does "in the course of" mean?

An invention made or conceived in performing, or as a result of performing, the work required by a contract is made or conceived "in the course of" that contract. That would be true even though the invention was not specifically sought in the terms of the contract. An invention is made or conceived "under" a contract when it is made or conceived during the life of the contract and the invention is, in whole or in part, specifically provided for by that contract.

This also suggests that a connection exists through the specific work assigned performed (to invent, or to solve and thereby invent), and not the general area of application.