Which default action wins the hard constraint?
Only reviewed-release (risk 0.05) and abstain (risk 0) are authorized and within the limit 0.1; fast-release is unauthorized and risky-authorized has risk 0.2. Of the two, reviewed-release has the higher reward, 5 against 0.
Can risk_limit = 0.25 authorize fast-release?
No. Its current permission remains false, and its risk 0.3 also exceeds the limit 0.25.
Why does transfer choose wait despite negative reward?
It is the only authorized action satisfying risk<=0.