Slow to perform terraform state refresh with 46 servers

I managed to find the rate limiting factor was the number of maas server workers.

From How to manage regions, the default was 4 workers which matched the results of my experiment above.

After increasing workers to 16 on the maas server with 16 cpus, using parallelism of 10 gave a time of 2m28s (compared to the previous value of 6m17s)