Decision to refactor and use time-series collections

Hello, I’m looking for guidance on whether or not a proof-of-concept / refactor is worthwhile to leverage timeseries functionality.

The current data model has a single collection which contains 3 different time based documents (let’s call them A,B,C) which are sourced from real time instrumentation. In terms of spread, documents of type C constitute 90% of the data. The collection has an index on the timestamp

Typically a given weekday will see > 10000 records added which correlate to 126 unique locations

The most common report generated from this data finds the latest ‘A’ record for a given location and uses the timestamp of A to find all future dated records of B and C (for the same location) and aggregates a result depending on what the report requires.

Specifically I’m curious whether separating each of the document types into their own timeseries collection and changing the aggregate to use $lookup (eg: A.aggregate (lookup “B” {}, lookup “C” {}) will provide any performance improvement or if, in fact, it could be less performant.