Записывать данные только в строку csv, не теряя предыдущие данные?

Это один из способов получить желаемый результат. Встроенные комментарии должны помочь в мыслительном процессе.

# Import-Csv here instead
$sourceCsv = ConvertFrom-Csv @'
Underlying,AllBlue,AllRed
AUD,1/5/2024,
BRR,,
GBP, 3/10/2024,
CAD,,
'@

# Import-Csv here instead
$newCsv = ConvertFrom-Csv @'
Underlying,AllBlue,AllRed
AUD,,
BRR,4/29/2024,
GBP,4/29/2024,
CAD,,4/29/2024
ThisIsALineToAppend,,4/29/2024
'@

# create a hashtable using the Underlying property as keys
$map = $sourceCsv | Group-Object Underlying -AsHashTable

$linesToAppend = foreach ($line in $newCsv) {
    # if the Underlying value of the new Csv exists in the old CSV
    if ($map.ContainsKey($line.Underlying)) {
        # get the object having the same Underlying value of the old CSV
        $value = $map[$line.Underlying][0]
        # if the AllBlue of the new CSV is not null
        if (-not [string]::IsNullOrWhiteSpace($line.AllBlue)) {
            # update the object of the old CSV with this new value
            $value.AllBlue = $line.AllBlue
        }
        # if the AllRed of the new CSV is not null
        if (-not [string]::IsNullOrWhiteSpace($line.AllRed)) {
            # update the object of the old CSV with this new value
            $value.AllRed = $line.AllRed
        }

        # go to the next iteration
        continue
    }

    # else, this Underlying doesn't exist in the old CSV
    # so its gotta be appended
    $line
}

$sourceCsv + $linesToAppend | ConvertTo-Csv

Это расширяет ответ @Santiago. Это позволит нескольким полям составлять составной ключ для каждой строки. Он также будет сравнивать все неключевые поля, не требуя их жесткого кодирования. Это несложно, если есть два (2) поля для сравнения. Гораздо сложнее, если нужно сравнить восемьдесят два (82) поля.

$NaturalKey = @('Underlying','New')

# Import-Csv here instead
$sourceCsv = ConvertFrom-Csv @'
Underlying,New,AllBlue,AllRed
AUD,9,1/5/2024,
BRR,8,,
GBP,7, 3/10/2024,
CAD,6,,
'@

# Import-Csv here instead
$newCsv = ConvertFrom-Csv @'
Underlying,New,AllBlue,AllRed
AUD,9,,
BRR,8,4/29/2024,
GBP,7,4/29/2024,
CAD,6,,4/29/2024
ThisIsALineToAppend,5,,4/29/2024
'@

# Compare the fields of both CSV files to ensure that they have the names.
# The fields do not need to be in the same order.
$SourceFields = $sourceCsv[0].psobject.Properties.name | Sort-Object
$NewFields = $newCsv[0].psobject.Properties.name | Sort-Object
if ($null -ne (Compare-Object -ReferenceObject $sourceFields -DifferenceObject $NewFields)) {
    Write-Error -Message 'CSV files do not have the same field names or number of fields."'
    return 1
}
# Only test for new values in the fields that are not part of the natural key.
$FieldsToCompare = foreach ($Field in $NewFields) { if ($Field -notin $NaturalKey) { $Field }}

# create a hashtable using the Underlying property as keys
$map = $sourceCsv | Group-Object $NaturalKey -AsHashTable -AsString

# Iterate over all rows of the new CSV file.
$linesToAppend = foreach ($line in $newCsv) {
    # Gather the fields comprising the natural key.
    $KeyValues = foreach ($k in $NaturalKey) { $line.$k }
    $Key = $KeyValues -join ', '
    # if the natural key value of the new Csv exists in the old CSV
    if ($map.ContainsKey($Key)) {
        # get the object having the same Underlying value of the old CSV
        $value = $map[$Key][0]
        # Iterate over the fields to be compared (all fields except those in the natural key).
        foreach ($Field in $FieldsToCompare) {
            # If the new field valus is not null or whitespace, replace the existing value.
            if (-not [string]::IsNullOrWhiteSpace($line.$Field)) {
                $value.$Field = $line.$Field
            }
        }
    } else {
        # else, this Underlying doesn't exist in the old CSV
        # so its gotta be appended
        $line
    }
}

$sourceCsv + $linesToAppend | ConvertTo-Csv