From my armchair here I thought something similar, I think I heard 1) that inserting into the same table has some speed limitations and 2) one of the authors ideas was to spin up 4 processes…

I wonder if you could insert into 10 different tables from 10 threads or processes then

insert into single_table (…) select (…) union.

No idea if insert select does the trick or not but you’re almost at the point of partitioning here. If the application called for it you could do some sort of poor mans partition in order to write all the data and then push some complexity into the read.

I also wondered what if any impact the ID primary key had on inserting at the level of frequency.

/armchair

If you define the virtual table would you even need to do the `INSERT`? Just query the virtual table instead of inserting into `t` and querying `t`.