Crypto Payment Confirmation System with Reorg Protection

Crypto Payment Confirmation System Development

Imagine a client pays an order in USDT on Polygon, but a minute later the network reorganizes — the transaction disappears. You already shipped the order, but the money never arrived. A reliable payment confirmation system is not just a hash check; it's a finite state machine with explicit state transitions and protection against all edge cases. Our implementation uses separate monitors for each network, a tolerance window to handle amount fluctuations, and idempotency at the txHash level. For example, on Ethereum PoS we require 12 confirmations (average 12-second block time), providing reliability comparable to bank clearing but 10 times faster.

We build such systems from scratch or integrate them into existing infrastructure. We rely on the EIP-1559 and Ethereum JSON-RPC API specifications for correct transaction processing. Operational cost savings on payment processing can reach $2,000 per month. The system typically pays for itself in 3–4 months. Turnkey delivery in 2–4 weeks. Contact us to discuss your scenario.

What Problem Are We Solving?

A naive implementation: receive hash → check amount → credit. It breaks at the first reorg, double spend, or when the user sends payment an hour after session expiry. Main pain points:

• Reorg: A block is abandoned, transaction disappears. Without status rollback, you credit non-existent funds.
• Floating point: Conversion via wei introduces rounding errors; user pays 47.50 USDT, but the system sees 47.499999.
• Exchange fees: Transferred amount is 1–2% less than expected.
• Session timeouts: Payment arrives after the time limit, and the address is no longer valid.

Each of these problems is resolved within a unified finite state model.

How Does the System Protect Against Reorg?

Reorg is a chain reorganization where a previously accepted block is replaced by another. On Ethereum PoS this is unlikely (1–2 block depth), on Polygon it's more common. Our approach: on every confirmation check, we fetch a fresh transaction receipt. If the receipt disappears, the status rolls back to DETECTED, the counter resets, and the monitor begins re-searching.

async function processConfirmations(paymentId: string) {
  const payment = await db.findPayment(paymentId);
  const currentBlock = await provider.getBlockNumber();
  const receipt = await provider.getTransactionReceipt(payment.txHash);
  if (!receipt) {
    await db.updatePayment(paymentId, {
      status: 'DETECTED',
      confirmations: 0,
      reorgDetected: true,
    });
    return;
  }
  const confirmations = currentBlock - receipt.blockNumber + 1;
  const isConfirmed = confirmations >= payment.requiredConfirmations;
  await db.updatePayment(paymentId, {
    confirmations,
    status: isConfirmed ? 'CONFIRMED' : 'CONFIRMING',
    confirmedAt: isConfirmed ? new Date() : null,
  });
}

Payments in CONFIRMING status are rechecked every N blocks — we never trust stale data.

What If the User Sends Less or More?

Due to fees and floating point, the transaction amount rarely matches the expected amount exactly. A sensible tolerance window solves this. Verification code:

function isAmountSufficient(
  received: bigint,
  expected: bigint,
  toleranceBps: number = 50
): 'exact' | 'underpaid' | 'overpaid' {
  const tolerance = expected * BigInt(toleranceBps) / 10000n;
  const min = expected - tolerance;
  const max = expected + expected / 10n;
  if (received >= min && received <= max) return 'exact';
  if (received < min) return 'underpaid';
  return 'overpaid';
}

On underpaid, the system notifies the operator; on overpaid (up to 10%), it accepts the payment and credits the surplus to the user's balance or generates a refund.

Payment State Machine

Each payment passes through strictly defined states: PENDING → DETECTED → CONFIRMING → CONFIRMED → SETTLED ↓ ↓ EXPIRED UNDERPAID / OVERPAID ↓ REFUNDED

Blockchain Monitor Architecture

Monolithic monitoring of all networks in a single process is a bad idea. We use a separate worker per network with an independent retry mechanism. Implementation for EVM networks:

interface ChainMonitor {
  network: string;
  start(): Promise<void>;
  stop(): void;
  onTransaction(handler: (tx: IncomingTransaction) => Promise<void>): void;
}

class EvmChainMonitor implements ChainMonitor {
  private provider: ethers.JsonRpcProvider;
  private watchedAddresses = new Set<string>();

  async start() {
    const activePayments = await db.query(
      "SELECT address FROM payments WHERE status IN ('PENDING', 'DETECTING', 'CONFIRMING')"
    );
    activePayments.rows.forEach(p => this.watchedAddresses.add(p.address));
    this.provider.on('block', async (blockNumber) => {
      await this.processBlock(blockNumber);
    });
  }

  private async processBlock(blockNumber: number) {
    const block = await this.provider.getBlock(blockNumber, true);
    for (const tx of block.transactions) {
      if (tx.to && this.watchedAddresses.has(tx.to.toLowerCase())) {
        await this.handleNativeTransfer(tx, blockNumber);
      }
    }
    await this.scanErc20Transfers(blockNumber);
  }
}

Confirmation Requirements for Different Networks

Idempotency and Duplicate Protection

One txHash must be credited exactly once. We use INSERT with ON CONFLICT DO NOTHING: if the same hash already processed, it returns an empty result.

INSERT INTO payment_transactions (payment_id, tx_hash, amount, block_number) VALUES ($1, $2, $3, $4) ON CONFLICT (tx_hash) DO NOTHING RETURNING id; 

Notifications and Webhooks

After transition to CONFIRMED — immediate notification to external systems via a queue (Bull/BullMQ) with exponential backoff. Direct HTTP call in the block handler would lose events on failures.

async function dispatchPaymentConfirmed(payment: Payment) {
  await eventBus.emit('payment.confirmed', {
    paymentId: payment.id,
    orderId: payment.orderId,
    amount: payment.receivedAmount,
    txHash: payment.txHash,
  });

  if (payment.webhookUrl) {
    await webhookQueue.add(
      'payment-webhook',
      {
        url: payment.webhookUrl,
        payload: {
          event: 'payment.confirmed',
          data: payment,
        },
      },
      {
        attempts: 5,
        backoff: {
          type: 'exponential',
          delay: 2000,
        },
      }
    );
  }
}

Process and Deliverables

1. Analysis — we dissect your business requirements, number of networks, tokens, refund scenarios.
2. State machine design — refine transitions, tolerance, confirmation thresholds.
3. Implementation — code monitors, handlers, webhooks, integration tests.
4. Testing — cover edge cases: reorg, underpaid, timeout, double-spend.
5. Deployment and monitoring — deploy in your infrastructure, set up alerts.

What is included in the result:

• Source code repository with launch instructions.
• API and architecture documentation.
• Database migrations.
• Load tests and simulation scripts.
• Support for 2 weeks after launch (extended support on request).

Order the development of a system for your project — we will prepare a detailed estimate within 1 day.

Timeline and Guarantees

Typical delivery time is 2 to 4 weeks, depending on the number of networks and business logic complexity. Pricing is calculated individually, but we guarantee transparent cost breakdown. We have been working with blockchain projects for over 5 years and have implemented dozens of such systems. We guarantee stable operation under a load of up to 10,000 transactions per hour.

Get a consultation: write to us, and we will evaluate your project for free.