To more reliably reproduce external-to-nodeos performance characteristics (such as disk IO, memory caching, etc) during live operation of the chain, it seems like it might be useful to replay blocks at their approximate real time interval. Otherwise when replaying 1000 blocks in a few seconds it's harder to know, for example, how disk writes behaved during the 500 original seconds those blocks arrived in, when instead during replay they're compressed down to being replayed in a handful of seconds where short term buffering may mask performance characteristics.
To more reliably reproduce external-to-nodeos performance characteristics (such as disk IO, memory caching, etc) during live operation of the chain, it seems like it might be useful to replay blocks at their approximate real time interval. Otherwise when replaying 1000 blocks in a few seconds it's harder to know, for example, how disk writes behaved during the 500 original seconds those blocks arrived in, when instead during replay they're compressed down to being replayed in a handful of seconds where short term buffering may mask performance characteristics.