Ambiguous interpretation of the CompletionList.isIncomplete field

[BaneShand]

Our team had different interpretations of the LSP specification regarding the CompletionList.isIncomplete field. We identified two possible client behavior scenarios:

First opinion:
If isIncomplete = true, the client should request new options when the text changes (new textDocument/completion call).
If isIncomplete = false, the client considers the list complete and does not expect new items from the server until the context changes. It also DOES NOT SEND REQUESTS until the context changes.

Second opinion:
If isIncomplete = true, the client should request new options when the text changes.
If isIncomplete = false, this may mean that the current list is partial, and the server will send the remaining items as they become available (for example, through additional notifications for the same request).

Question:
Which of these interpretations is correct? Can isIncomplete = false mean that the server will continue to send items for the current request, or does LSP not provide for this, in which case partialResultToken should be used?