I will argue here that Dreaming (REM) sleep entails that Short Term Memory (STM) remains active, while Executive Function is disconnected from Long Term Memory (LTM).
Executive Function refers to our explicit awareness of choice in what we 'think about' (and plan to do) - Executive Function can control Long Term Memory (LTM) such that it can initiate the activation of LTMs into STM.
In other words, Executive Function can direct thought by causing retrieval of Long Term Memories, and can activate them in Short Term Memory - and it is this choice and retrieval process which is suppressed in Dreaming sleep.
The result of suppressing Executive Function is that we do not control our dreams; instead our dreams happen, our awareness 'observes' what happens, and we try to make sense of it all.
Short-term memory (STM), also - in a rather expanded concept - called Working memory (WM), is considered to be the process - lasting some seconds - by which material is held in awareness; perhaps in the form of symbolic 3-D patterns of activated neurons, which can actively be combined, sequenced and manipulated. This is reckoned to happen in the (probably lateral) pre-frontal cerebral cortex.
By contrast, Long term memory (LTM) is conceptualized as a structural change in brain circuitry - perhaps encoded in the form of sequences or patterns of facilitated pathways, or differential synaptic strengths.
J Allan Hobson (in The Dream Drugstore, 2002) suggests that the lateral pre-frontal cortex, hence STM/ WM, is suppressed or switched-off in Dreaming (REM) sleep - and he relates this to suppression of the aminergic neurotransmitter systems.
But I think this is only partly correct, since the phenomena of dreams are much better explicable if we suppose that STM remains fully active in dreaming sleep, but there is a disconnection between STM and LTM.
In the waking state, STM is capable of mobilizing and activating Long-term memories (to translate patterns of synaptic strengths into patterns of neural activation), in order to 'make sense' of Short-term memories using accumulated knowledge and experience.
But, I suggest that the STM remains fully active in Dreaming sleep, and 'doing its best' to make sense of the material being 'thrown at it' by activations of LTM that it no longer controls - thus, in dreaming sleep the STM is disconnected from LTM and does not have the ability to mobilize or control LTM.
If so, this contributes to convergent evidence concerning the specific function of Dreaming (REM) sleep in terms of 'testing' the short-term coherence and inter-consistency of variously-activated long-term memories, by stringing them together in various permutations in narratives.
If Dreaming sleep is a testing process, then it would make sense that the test would be of the nature of a cross-check between memories - rather than following the memories in the same order and with the same specific meanings as they had when they were stored.
This would imply that for the testing process to work, the order and manner of activation of Long Term Memories should not be under the control of Executive Function - because EF would tend to elicit memories in the temporal sequence and semantic (meaning) categories they were originally stored.
The principle upon which these LTMs are elicted in Dreaming sleep is not precisely known.
Clearly there must be a principle or principles, since nature is never random. Something of these principles is understood insofar as a 'dream logic' is recognized, based upon associations, or similarity of elicited emotions (what one is 'reminded-of').
So dreams are made from these disparate LTMs, which are activated simultaneously and in sequence according to the poorly understood principles of dream logic - and Short Term Memory 'inspects' these memories, and tries to synthesize them into meaningful wholes within the constraints of the (several to tens of seconds) timespan and finite processing power of STM.
In other words, Dreaming sleep constitutes a check of the coherence and inter-consistency between LTMs from different periods of time, and with different subject matter and meaning; on the assumption that 'reality' must be mutually coherent, therefore any two memories on any subject from any point in personal history, should be compatible...
And if they are discovered to be not compatible, then these memories need to be 'edited' to make them compatible.
And that checking and editing is what is happening during Dreaming sleep, as partially sampled during dreams.