[bortraws]

Your 'real-time' solution would need PT to look ahead for the plugin processing costing extra CPU overhead.

Automatically shifting the audio earlier in time when finished with recording is the easiest and most CPU efficient way of doing it. And it can be done sample accurate so no problems there. The delay could be 4000 samples or even seconds. It wouldn't matter. This only gives you an initital delay when starting to record and no more than the amount of samples from the input buffer as your monitoring delay. This releases the strain on the processor giving you all that extra power for plugins. Best solution IMO.

But we'll have to wait and see how it's done.