Knowing that very low percentage of E. coli mRNA is polyadenylated, perhaps random hexamer should also be used to generate even the 1st strand of cDNA, instead of oligo (dt) primer?
Well in this case you can use both but as you said that a very low percentage of E. coli mRNA is polyadenylated so mu suggestion to you is to use random primers (mostly hexamers) which can hybridize with or without polyA for cDNA synthesis.
It actually depends on what genes you are looking for.
For example, if you are only seeking the Ecoli genes, it is highly recommended to use random hexamer as you clarified above.
But if your are looking for the cellular response to bacterial infection (e.g. immune response genes), make another cDNA preparation but by using (oligo dt) as it is much more easier to proceed with (on cDNA, qPCR levels).
Each time (either in random hexamer or in oligo dt) make a qPCR run for GAPDH or any other housekeeping gene as an internal control for the calculation of your relative gene expression.
Using random hexamer seems better option in this case. However, if you wish to use Oligo(dT) as a primer for first strand cDNA synthesis then it is necessary to enrich your mRNA preparation with biotinylated (dT)20-25. Please refer to: https://www.pnas.org/content/pnas/89/16/7546.full.pdf for more details.