cluster.fork() fails when called in repl (cluster.settings.exec undefined)

[bpstrngr]

Running v14.4, I launched my script with cli dynamic import:
node -e 'import("./seed.js").then(({open})=>open());'
and I get the following error on fork process creation:

TypeError [ERR_INVALID_ARG_TYPE]: The "modulePath" argument must be of type string. Received undefined
    at validateString (internal/validators.js:121:11)
    at fork (child_process.js:63:3)
    at createWorkerProcess (internal/cluster/master.js:132:10)
    at EventEmitter.cluster.fork (internal/cluster/master.js:167:25)
    at file:///media/fuse/drivefs-fb63ac1d072197a08fd8e42d246fd29a/root/www.blikpatrik.net/Blik_2020_seed.js:40:58
    at Array.map (<anonymous>)
    at file:///media/fuse/drivefs-fb63ac1d072197a08fd8e42d246fd29a/root/www.blikpatrik.net/Blik_2020_seed.js:40:41

in internal/cluster/master.js:132,
createWorkerProcess(id,env) function passes cluster.settings.exec down empty to fork(modulePath)
after cluster.setupMaster() has it set from process.argv[1]
(which is empty due to --eval usage)
at internal/cluster/master.js:57.

I see running code dynamically as an equally reasonable use case
for forking processes so i suppose this is not intended.

I wonder if import.meta.url could be utilised for a solution, at least as an optional argument.

Thanks for the attention!

only related issue i found:
#27039

[bpstrngr]

The solution for me in the meantime has been to set a mock entry point before forking,
which a bit surprisingly doesn't influence the childprocesses
launching with the correct initial command so it works fine.
if(!process.argv[1])
process.argv[1]=import.meta.url;