Biological plausibility for a delay connection of about 100 msec

Dear all
I am using the delay method (implemented using nodes) for forwarding neutral information in SPA (parallel delay of 500 values: S-pointer).

The problem in my large-scale model for sentence comprehension is as follows:
In my syntactic processing module of my sentence comprehension network the assignment of syntactic labels to words takes about 100 msec. But then I need to associate these grammatical word labels (lemmata) with concepts of words (which are retrieved faster from the mental lexicon).
In order to be able to do a correct role assignment, i.e., a correct semantic word association with the syntactic word labels (online) I need to delay the semantic information of the words by about 100 msec.

My question:
Because I want to build up a biologically plausible model , my question is:
How can I implement a delay processing module for delivering delays of about 100 msec for neural information stored in about 500 ensembles (S-pointers) in a biologically plausible way?
(I think, the solution of implementing a delay routine with nodes is not biologically plausible)