I'm wondering if there is consensus in what effects "turning subpatches on/off with swtich~ objects" has, mostly in terms of cpu-spikes.
I'm pondering if "turning off" modulation sends/receives, controlled by a modwheel for example, would be computationally beneficial. This would mean that the switch~ could trigger fairly often, and be "on" for short durations.
So the question is:
Do switch~'es operations cause spikes in cpu-usage in a way that would make this kind of thing counterproductive?
A lazy secondary question: Is it reasonable to expect switches to respond smoothly in a real time application? Will there be a delay noticeable delay or are they pretty much instantaneous? (a 64s block or two is to me fully acceptable).

Cheers!

Have you given this a try and got CPU spikes? I don't think spikes occur when using [switch~].
As for the second question, it's super responsive. Since it is being controlled in the message domain, expect it to be enabled at the next sample block.