I am using in Development phase, so unable to see not much data processing, but these questions in my mind are for future concerns, and also how much to depend on workers compare to how much to depend on API direct KV access. Thanks for your time to answer the concerns.
I am still not clear. In the above question I used may be wrong example or wrong words. For a second ignore the 50ms limit (say, I was able to buy larger time for each slot.) The question is the 330 seconds above will it be completely charged. There is a concept like the FRAMETIME (time frame?), where the CPU is reading the image, but the actual processing is only 30 ms. Will 1) and 3) above included in the CPU time or not (the reading image from KV and the streaming part to client.)
Or if the 50 ms stays, will my worker crash because the total length is 330 ms (but 330 ms is not completely processing - there is some waiting for data phase).