In this case it probably works most reliably with around this number of As, and less/more would decrease reliability, butnitmwould still be functional.
With more As you risk running into the upper limit on sequence length for the virus shell and with less you run the risk of quicker degradation and not enough expression
The first point being only a concern if you are a virus. If instead you are engineering say, an rna vaccine, you could package your transcript into lipofectamine instead of a coronavirus envelope ;) but point taken, there's always an upper limit.