I've noticed a few crash cases when developing jsonnet locally:
sufficiently gnarly jsonnet can cause the lookup to infinite loop and blow the stack
IntelliJ doesn't always consistently update the version index by 1, leading to a crash on the diff update. typically violations are just skipping a single version number and as far as I can tell, the LSP state would have been fine if it had just accepted the new update -- these cases appear to have happened during git operations in which IntelliJ is sending the whole file as an update anyway.
I've noticed a few crash cases when developing jsonnet locally: