I frame my response in terms of an analogy with reading theory. A text will provoke multiple target concepts according to whether the recipient is more or less richly endowed with appropriate background schemata (relevant prior mental experience). For instance, a subject lacking in such experience might merely decode 'words' (perhaps as a sequence of simple dictionary definitions), whereas someone who has a richer textual repertoire will (or may) experience multiple targets being stimulated, not only concepts, but feelings, associations, cultural references, aesthetic associations or religious and political presuppositions. These domains will be activated simultaneously, but it takes analytical skill to isolate and articulate them separately, i.e. to make them accessible. Much of cognitive linguistics has an awkward and under-acknowledged relationship with the technologies of high literacy.

From my experience with multimodal data, as shoen by Müller & Cienki (2009), for metaphoric mappings, you can have: the same source and target in two modalities (varbal and gestural); the source in one modality (gestural) and the target in the other (verbal); and, eventhough the authors predict that you can have the same source and different targets, it wasn't attested so far by data collection. Maybe this shows us a cognitive constraint for having such structures.

Nevertheless, from the point of view of Mental Spaces Theory, it is possible to consider mental models such as megablends (Fouconnnier & Turner, 2002, chap. 16), megablends, wich result from "blends of blends" (the blending works as an input space for another one). Nevertheless, all this processes have also constraints.

Finally, in general, we have several models were developed beyond the source-target domain proposed on the initial years of the Conceptual Metaphor theory - i.e books like Forceville 2009 (and more recent papers), Müller 2008 (and more recent papers), Nuñez & Sweetser very relevant paper from 2006, Kok and Cieki 2016, about neuronal issues). But in all of these models, there are not only neural constraints, but also cultural specificities that prevent us from talking about "universals", even when we talk about primary or basic metaphors (that can be found in several cultures).

I will cut your query into multiple bits.

(i) It is a fact that a source domain (SD) may structure multiple target domains (TD). For instance, the SD of BUILDING in my culture will structure TDs as diverse as theory, career, economic system/company, life, educating children, imagining, etc. (This can be read about Tunisian Arabic in Kovecses Metaphor in Culture, 2005: pp. 77-78).

(ii) However, it is not clear to me how in the neural theory of metaphor "multiple target domains should be activated simultaneously." I cannot think of Feldman (2006) and Lakoff (2012) having said this. For instance, if I take the metaphoric expression, "Here is an idea that I'd like to plant in your mind" (Lakoff & Johnson, 1980/2003: p. 47). Does it mean that "planting" or "plant" activate here more than the TD "idea"? If this were the case, the situation with metaphor will be chaotic, with a SD triggering the TD in question along with other domains.

(iii) Perhaps Junwei means by "multiple target domains" multiple properties of the SD as applicable to the TD in context, thus constituting the mapping. For Lakoff (2012), a Neural Theory of Thought and Language "assumes spreading activation and inhibition along existing neural pathways, and the concept of 'best-fit'" (p. 779). Feldman (2006) accounts for the concept of best-fit, noting that "the brain is inherently a best-match computer" (p. 13).

I hope this will be of help.

Em meu entendimento, quando existe tal afirmativa na teoria neural da metáfora, o que está em jogo é a múltipla possibilidade de acionamentos advindos de domínios-fonte determinados. Quando eu imagino, por exemplo, a ação de pegar algo, o córtex pré-motor é acionado e as múltiplas possibilidades de acionamentos podem acontecer, porém a restrição ocorre quando associamos o domínio-alvo específico ao domínio-fonte, como, por exemplo, pegar uma caneta, ou pegar uma ideia. Independente do nível de abstração da formação metafórica, o que está sendo construído é uma emulação de ação, o que nos faz compreender que o domínio-fonte já é restringido desde a sua base de formação, pois caso contrário entraríamos na concepção seriada de acionamentos neurais ou, ainda, remontaríamos aos princípios gerativistas que partem de uma linearidade.