Filter on Time-Series with &lte on timestamp and reverse index is too slow

Working with Data
, ,
1

Hi there, I have this time series collection:

{
  "t": {
    "$date": "2022-04-20T18:20:00.000Z"
  },
  "c": "5b18b3a6102d58",
  "_id": {
    "$oid": "635e639985470d444ea26fbb"
  },
  "v": 13037430
}

I have created the following indexes: t_1_c_1, t_-1_c_1, c_1_t_1, c_1_t_-1

Now, when I run this query:
filter: {"c":"afadf7cfb8cefa7","t":{"$gte":"2024-08-01T00:00:00Z"}}, sorting: {t: 1}
it is resolved within 1 second, but when I do this:
filter: {"c":"afadf7cfb8cefa7","t":{"$lte":"2024-09-01T00:00:00Z"}}, sorting: {t: -1}
it takes minutes!!!
When I run explain in MongoDB Compass, it shows that the ixscan was quick (app 4s) but then fetch takes easily 2 minutes but had seen 5 minutes easily. :confused: The best I got to was about 14 seconds!

I tried to use hints but without a real success. It shows that it uses correct index but the data are still being retrieved very slowly.

I have never seen such a behaviour from a DB before. :frowning:

Any help? Thanks!