Tests have shown inconsistent treatment of the If-None-Match: * header.

Vague. [Citation needed]

I believe there is consensus that Solid Servers must support this feature based on the issues discussions in #40 and #108.

Already in the Protocol:

A data pod MUST implement the server part of HTTP/1.1 Conditional Requests [RFC7232] to ensure that updates requested by clients will only be applied if given preconditions are met.

Moreover, for clients the Protocol includes:

A Solid client MAY implement the client parts of HTTP/1.1 Conditional Requests [RFC7232] to only trigger updates when certain preconditions are met.

Hence, neither the original or the proposed text is necessary.

As for:

I also think we cannot put such a requirement on the client

The requirement is not pushed to the client. Client behaviour is compatible with servers irrespective of the header in the request.

, also we can't test that.

Why? (Putting aside the fact that it is not the client that actually prevents undesired modification.)

Thus, I have changed the wording to make it a requirement on the servers.

Redundant.

I don't agree with the proposed change.

As mentioned, since server support is a MUST and client support is a MAY (as per #http section), no text is necessary. The purpose of the existing text in the Protocol was to (strongly?) encourage clients to use the header. So, I propose to change it to a Note - it is not a whole lot different than a MAY here but at least it is not redundant and still gets to put some emphasis on its use.