Executing many Dice calls simultaneously on a single machine causes race conditions

[caleb-johnson]

Environment

All supported

What is happening and why is it wrong?

Dice uses a time stamp to name its shared memory artifacts, which causes race conditions and unexpected behavior when spinning up more than one job within a second.

How can we reproduce the issue?

Execute many Dice jobs in parallel as quickly as possible (Can use Ray/Qiskit Serverless)

Traceback

No response

Any suggestions?

Document this behavior so users can build a 1+s delay into their workflows.