I have read the articles and understand the "Parameter Sniffing" issues and some of the methods that may prevent / reduce this.

In my case I am still unsure why an entirely new plan was created, rather than just use the original less than optimal plan for the increased result set.

If the statistics are getting outdated on some tables then I may have to updates statistics on one or two big tables during the day - is this a possible practice to do or is it classed as bad practice / last resort?