Remove rows with symmetric values by a combination of columns

I have a dataframe and I want to remove rows that have a symmetric value in a column based on a combination of other columns. In essence I want to remove the refunds from my sales dataframe.
My initial dataframe looks like this:

df <- data.frame(
  clientID = c(101, 101, 102, 103, 103),
  transactionID = c(1, 2, 3, 4, 5),
  date = as.Date(c("2023-05-01", "2023-05-02", "2023-05-03", "2023-05-04", "2023-05-05")),
  productID = c("P001", "P002", "P003", "P004", "P005"),
  QTY = c(2, 3, 1, 5, 2)
)
refund_rows <- data.frame(
  clientID = c(101, 102, 103, 101),
  transactionID = c(6, 7, 8, 9),
  date = as.Date(c("2023-05-07", "2023-05-06", "2023-05-08", "2023-05-09")),
  productID = c("P001", "P003", "P005", "P006"),
  QTY = c(-1, -1, -2, -5)
)
final_df <- bind_rows(df, refund_rows)

I want my final dataframe to look like this:

clientID transactionID date productID QTY
101	2	2023-05-02	P002 3
103	4	2023-05-04	P004 5
101 9   2023-05-09  P006 -5

How can I do this in R?

I tried the following but the problem is I am left with the transactionID = 9 which should be negative QTY

final_df <- data.frame(
  clientID = c(101, 101, 102, 103, 103, 101, 102, 103, 101),
  transactionID = c(1, 2, 3, 4, 5, 6, 7, 8, 9),
  date = as.Date(c("2023-05-01", "2023-05-02", "2023-05-03", "2023-05-04", "2023-05-05", "2023-05-07", "2023-05-06", "2023-05-08", "2023-05-09")),
  productID = c("P001", "P002", "P003", "P004", "P005", "P001", "P003", "P005", "P006"),
  QTY = c(2, 3, 1, 5, 2, -1, -1, -2, -5)
)

refund_rows_new <- final_df[final_df$QTY < 0,]

refund_rows_abs <- refund_rows_new %>% 
  mutate(QTY = abs(QTY))

final_df_new <- final_df[final_df$QTY > 0,]

final_df_new %>% anti_join(refund_rows_abs, by = c("clientID", "productID", "QTY"))

My first idea would be to group by clientID and productID and filter based on the sum of QTY.

final_df %>%
  group_by(clientID,productID) %>% 
  filter(sum(QTY)!=0) 

  clientID transactionID date       productID   QTY
     <dbl>         <dbl> <date>     <chr>     <dbl>
1      101             1 2023-05-01 P001          2
2      101             2 2023-05-02 P002          3
3      103             4 2023-05-04 P004          5
4      101             6 2023-05-07 P001         -1
5      101             9 2023-05-09 P006         -5

however this gives a different result from what you have requested. because client 1 bought 2 of product P001 and got a refund for 1.

So if you want to omit those entries you could do something in the lines of:

final_df %>%
  group_by(clientID,productID) %>% 
  summarise(QTY=sum(QTY))%>% 
  filter(QTY!=0) %>% 
  left_join(final_df)

  clientID productID   QTY transactionID date      
     <dbl> <chr>     <dbl>         <dbl> <date>    
1      101 P001          1            NA NA        
2      101 P002          3             2 2023-05-02
3      101 P006         -5             9 2023-05-09
4      103 P004          5             4 2023-05-04

and omit rows containing NA

final_df %>%
 group_by(clientID,productID) %>% 
 summarise(QTY=sum(QTY))%>% 
 filter(QTY!=0) %>% 
 left_join(final_df) %>% 
 na.omit()

Joining with `by = join_by(clientID, productID, QTY)`
  clientID productID   QTY transactionID date      
     <dbl> <chr>     <dbl>         <dbl> <date>    
1      101 P002          3             2 2023-05-02
2      101 P006         -5             9 2023-05-09
3      103 P004          5             4 2023-05-04

giving you the desired result, however this can be dangerous (because of the drop of client 1 and product 001).