For a *quantitative* real-time PCR, the amplicon size would be too large. But you just want to distinguish products of different sizes (see their precence, not quantify them). This is fine. An elongation time of 1min should be enough, though.

For question B there is no clear answer. Very long products might not be distinguishable by melting curves. For shorter products (~200bp) some 10% difference in size should be detectable by a shift in the melting temp (also depending a little on the actual sequence).

If the products are too long, you might use several exon-specific primers. Instead of distinguishing products by different melting temps you just see wheather or not a primer amplifies a product.

Another option was to design some exon-specific hydrolysis probes. They can be "multiplexed" in a single PCR. If you have to analyse the same gene in really a lot of samples, this would be worth a try, and it will be probably more cost-effective than using exon-specific primers in individual PCRs.

Regadring your question C: There are one-step rtPCR kits for real-time PCR available.

Note: real-time PCRs often contain higher concentrations of Mg++ as compared to conventional PCRs, this might reduce the specificity of the priming. You might have to use a different annealing temp. However, a robust system works always ;)

Yes, 670bp is far too big for realtime PCR. You go for amplicon sizes between 80 and 150bp (we usually tried to be between 80-120bp) for this.

Why do you need such big amplicons?

Distinguishing the products by melting curve depends on the base composition of the PCR product. I am not sure, if this works reliably.