Upsampling On Render/Playback Question

In general upsampling is favoured for synths, especially if doing any resampling. Every note produce aliasing artifacts.
Depending on material it's more obvious improvement or not.
 
So general idea is if plugins supports the doubled sample rate - you can use it in Sonar upsampling without having to run the full project in higher sample rate and everything that goes with that. More disk space, higher load on computer etc.
 
Upsampling is really picking the cherries in the Cake. But there are plugins like Metaplugin that allow it too.

There is little risk in upsampling a plugin's input, other than the remote possibility that a given plugin can't deal with it properly. You could probably upsample everything and not see a downside beyond the extra CPU consumption. But you might as well start by ignoring effects that you know won't benefit from it.
 
As a general rule, quality plugins only need upsampling if they are capable of creating harmonic distortion: primarily dynamics processors and distortion plugins. That includes limiters, compressors, gates, amp sims, tape sims, harmonic exciters and tape delays. But it isn't always obvious whether an effect introduces distortion or not, or whether the plugin deals with it internally. Filters and filter-based effects generally don't add distortion, and therefore don't need upsampling.
 
Even distortion plugins don't necessarily benefit from upsampling. Many do it internally anyway, or employ other techniques to avoid aliasing. High-end limiters such as Ozone or Pro-L, for example, simply do not alias, even at very aggressive settings. Even the much-maligned Boost11 doesn't alias easily.
 
If you really want to build a list of plugins that would benefit from upsampling, you can test your own collection yourself. Create a project with a single audio track and import or generate a series of sine wave pulses into it, and insert an instance of SPAN. Then stick in various plugins (in front of SPAN) and try to make them alias. What you're looking for is frequencies in SPAN's display that are not harmonically related to the fundamental test tone.