In any case, since 1000TPS is bandwidth limited, we dont really have any CPU bottleneck issues, unless you actually want to calculate crypto functions in the script instead of pushing it down with the script embedded in the AM
Each node (perhaps other than "light nodes") would have to *do* the crypto calculations as part of running the script otherwise once again you can just be tricked.
Again CPU may not be the main problem - but the size of the scripts would for sure be bigger than standard NXT txs are plus the VM "state" would also need to be stored somewhere (presumably as an AM - although currently that limits your persistent state to a paltry 1000 bytes).
OK, you confused me. If a client precalculates all the big calculations that is needed and pushes that onto the blockchain, then I dont think any node has to do any big calculations. They all can just use the data in the AM that the script reads in.
I think you are thinking about the scripts as running big huge programs, but I am thinking of them as doing stuff like check for data in AM, do a little work, save state in AM, done.
All the stuff like sending emails, sending DOGE, etc. would be done on the forging node's services modules. Something like that. Scripts are more like OpenCL kernels. Just the 1% that has to be done in the script as it gets a guaranteed runtime context on decentralized set of servers.
Just check to see if an email came in (via looking at input AM) and then triggering some external action with output AM.