It would depend on the expression levels of your templates or genes of interest. Do you know what Cts you were getting with what you had used before, i.e. with 200ng input for RT gave Ct of 20 for reference gene and 25 Ct for gene of interest(s)? Then you could work out by how much the Cts will shift since PCR is a doubling process, i.e.your Ct of 20 for the reference gene would become 21 if you had used 100 ng input (in liue of 200ng).

You will be at least needing 1ug of concentration in order to obtain 20ul of product. So based on your concentration ranges. dilution with rnase free water can be choose as an option. but always have a watch one levels of amplification.

I am not entirely sure why you are doing repeated RT reactions on the same RNA: why not simply make one (large) batch of cDNA from your RNA and then use that as your stock? Repeatedly dipping into your RNA is not only bad for your RNA (repeated freeze-thaw isn't the best for RNA integrity), but also your data, as your cDNA synthesis efficiency may vary each time.

I always prefer to use two-step protocols (make lots of cDNA, store that cDNA, use aliquots of the same cDNA stock for individual subsequent qPCRs) rather than one-step methods (make a small amount of cDNA and then use that cDNA immediately for qPCR).

Diluting your RNA won't actually do anything except perhaps marginally improve your pipetting accuracy. It won't give you more RNA. It's like making a cake when you need two eggs and only have one: diluting your egg with an equivalent volume of water doesn't change the fact you only have one egg.

Honestly, I would simply use ALL your remaining RNA to make one batch of cDNA: 8ul of a 20-40ng.ul-1 stock is a pretty decent yield, and all of your (highly labile, degradation-prone) RNA will then be in the form of analysis-tractable and more stable cDNA.