Confluence Overdraft & NSF PolicyRisk & Fraud

Task prompt

Hey, compliance just asked us to put together a quick brief on how we handle overdraft and NSF situations. Can you dig through our Confluence docs, the engineering tickets in Jira, and the warehouse to answer a few things? Some of this is only written down on the engineering side, so don't stop at Confluence. First, do we even offer overdraft protection — can customers spend below a $0 balance? I've seen some negative balances show up in our monthly account snapshots, so I want to understand how many there are and what's going on with those if we don't offer overdraft. For that count, stay in rds.monthly_account_snapshots and go off the ending available balance — count the snapshot rows sitting below zero on that column, not the distinct accounts behind them. Also, I need to understand our NSF handling for transfers. When we detect NSF from a high-risk counterparty institution, how many healing attempts does the system allow before it cancels? And for paycheck automations — if a PaydayInstance batch has 4 transfers but 1 of them would cause NSF, how many of those transfers actually go through? The paycheck planner spec documents how a payday batch is evaluated when the money isn't there, so answer that one from what the spec actually says the system does rather than from what you'd expect a bank to do. Give me the answers as: overdraft_offered (true/false), negative_balance_count_approx, nsf_healing_attempts, payday_nsf_transfers_executed.