The timestamp is, just like the price, data that comes from the exchange. It does not change based on how much the data provider is lagging. I actually did this experiment yesterday and only downloaded the tape for comparison today (historical data download). EDIT: Even if the data provider's latency was in there, it would be *at most* 1-2ms based on the speed of the cancels.
I don't have evidence, but I suspect that sending through SMART will be slower even if you disable dark pools. Having them to hold your order on their servers is probably worse than you holding them yourself, especially if you are colocated.
I just tried again in the premarket and it seems you are right, disabling dark pools made no difference. I actually got a submit time that is quite a bit worse than what I got yesterday (about ~2x on all stats). Seems quite likely now that the Good After Time system is somewhat flaky. I would like to test holding them myself but I am not set up near IB's servers yet so my own ping time (~100ms) would corrupt the experiment. That will have to wait until I get a dedicated server setup.
It depends how it is implemented. Could be implemented as a stand-alone module outside of the main processing. But even conceptually, you could say that PM margin involves some more computation. Most non-latency-sensitive clients don't put in the effort to measure these things. If they don't catch it by eye, it must be OK.
How is your experience using AWS in Virginia? Do you experience freezes caused by hypervisor as mentioned in the following post? I am evaluating different VPS providers and AWS is the one I like at the moment. Thanks for sharing https://www.elitetrader.com/et/threads/the-best-vps-for-automated-trading-to-ib.322975/
no freezes on c5 or x1 and I am watching it closely... all logs record time with milliseconds there are some 2-5ms delays here and there but nothing dramatic will move to dedicated server in NY eventually to shave off 5-7ms