2026.2.x Elevated Postgres Connections (1.5-3x from 2025.12.4)

### Describe the bug

While upgrading from 2025.12.4 to 2026.2.1, I'm seeing a 1.5-3x increase in postgres connections. This causes more load on the postgres servers and requires that they have higher resource caps for the raised connection counts. I had to raise my allowed connection count up from 200 to 500 to get 2026.2.1 to work.

With my setup on 2025.12.4 I rarely passed 100 connections in the last 30d. On 2026.2.1 its anywhere between 150-300. I nearly hit my 500 connection cap at one point during the upgrade process.

Although everything seems to work 'fine' with the raised postgres connection count and resources, I'm wondering if there's a known, significant change that is causing it.

### How to reproduce

Upgrade an instance from 2025.12.4 to 2026.2.1 and notice the shift in the `cnpg_backends_total` metric when using cnpg. Other db operators or instances will have a different metric for connection counts

### Expected behavior

Similar postgres connection counts to the previous version

### Screenshots

##### Past 30d

##### Past 24h (12h since upgrading)
ignore the two small dips in connection counts near 19:00; I was testing using a pooler. In the past I've had issues with a transaction pooler and the pending task count getting out of hand, so I decided against it (again).

### Additional context

_No response_

### Deployment Method

Kubernetes

### Version

2026.2.1

### Relevant log output

```shell

```