Github user cestella commented on the issue:
Yes, the caching semantics are the same as the enrichment topology; namely
that we cache all stellar expressions.
You can, however, turn caching off in the parsers in the case that you have
a nondeterministic stellar expression. I'd like to be able to augment stellar
function annotations to note expressions which are non-deterministic to avoid
caching in those cases.
Do you think we should turn it off by default?