Thanks @Hannah for the hints! I will check them out and report back here.
One thing that I could think of would be that RDS being on a t2.micro instance type might use up its cpu credits somehow and that’s why some dbase calls get refused (maybe due to the backup/maintenance processes being cpu intensive).
I also get a spike in index.php time in NewRelic, but unsure if it’s Ghost’s index.php or WordPress index.php.
Digging into NewRelic a bit I found these DB requests generating issues:
Anyways, will check the knex config and see if tweaking that to 0 as you recommended makes this going away.