I am about to alter about a 100 inherited tables to change the FK
reference to point to a more compact and efficient table representing
the PK for the relation.

Each FK drop/add will potentially take several minutes, so I'd like to
break it up over time to keep the main application from being totally
useless for several hours.

My question is can I implement it as a single slonik script with 100
EXECUTE SCRIPT commands and some "wait" between, or do I need to
implement 100 slonik scripts and execute them within a shell script
loop and appropriate wait in between.

That is, will the slonik command lock the whole set for the duration
of its execution, or will it lock the set per EXECUTE SCRIPT line?

Perhaps my best strategy will be to run this script on my replica node
only, swap master, then run it on the old origin using EXECUTE ONLY ON
parameter.

Thanks for any advice.