FTDC metrics per pool currently looks like this:

"NetworkInterfaceTL-TaskExecutorPool-0" : {
    "poolInUse" : 3,
    "poolWasUsedOnce" : 0,
    "poolConnUsageTimeMillis" : 10,
    "ip-10-128-56-199.ec2.internal:20040" : {
        "inUse" : 0
    },
    "ip-10-128-56-199.ec2.internal:20041" : {
        "inUse" : 0
    },
    "ip-10-128-56-199.ec2.internal:20042" : {
        "inUse" : 0
    },
    "ip-10-128-56-199.ec2.internal:20043" : {
        "inUse" : 3
    }
}

Each host currently has metrics for inUse, while each pool also has a poolInUse and a poolWasUsedOnce. We could add a wasUsedOnce metric to FTDC for each host, and this assert.soon could depend on the host's wasUsedOnce instead of a pool-wide poolWasUsedOnce.

The easier fix here would be to just relax the assert.soon and add a comment considering it's possible that the connection servicing {find:"shards"} or other queries may add 1 or more to the poolWasUsedOnce connection count.