Background

While PBWM has been proposed as a computational neuroscientific model of working memory, it has the following "unnaturalness."

While it assumes that the basal ganglia (BG) gate sensory input, they are supposed to gate the cortical output-thalamus loop.
cf. Benarroch, E. (2008) The midline and intralaminar thalamic nuclei Fig.2
While it assumes working memory is retained during NoGo, the default state for the BG, if this is the case, then working memory retention becomes the default state.

So an alternative model is proposed below.

Summary

Working memory retention is an action selected by the PFC.
Working memory retention, like any other action, is selected by the PFC-Th loop controlled by BG (the PFC-Th-BG loop) for execution.
The selection of working memory retention, like any other action, is exclusive, and only one is executed by the dlPFC at a time.
The working memory is retained for a certain amount of time due to the activity of recurrent (e.g., bi-stable) neurons in the PFC, and retaining it for a longer period of time requires to be "reselected" by the PFC-Th-BG loop.
The short-term memory (4) enables the dlPFC to hold multiple items in working memory.
The selection of dlPFC is also the selection of "attention".
The output from the activated dlPFC region is projected to the medial temporal lobe (around the hippocampus), from which relevant sensory information can be read. Information from the what and where pathways are integrated around the hippocampus, and hippocampal functions such as replay can also be used to access past information.
The prefrontal cortex uses the information stored in the working memory and the sensory information read out to predict actions and execute them to solve tasks.

Fig.1
The PFC-Th-BG loops for the premotor and motor cortices are omitted.

Neuroscientific validation and simulation