IBM webMethods Hybrid Integration

Are you saving the pipeline or using logged fields. If you save the pipeline then it will store it as a blob. This is both costly and of no use to you, as you cannot use it for tracking. 
if you use logged fields then it will simply save a key value pair for each attribute you checked. Cost is minimal and you can use this table to the find the transactions that you want. 

also don’t forget the service setContextID, which also allows you to set an explicit business identifier for your transaction and all subsequent steps. 

You have kind of answered your own questions. Audit DB isn't ideally suited for a production setup where you want to track data from a high volume call. It has never performed really well for the reasons you mentioned. There are a couple of alternatives:

1. Write your own  service to write to a store of your choice. You can call the service asynchronously or, if you need throttling, you can publish a doc and then call the service from the trigger.

2. Modern systems do this using OpenTelemetry traces and metrics. So if you have an APM tool, then you can go that route. I have done that at some customers using our OpenTelemetry package called Otelscope. But that is only relevant if you have an APM tool or are planning to use one.

@Rupinder Singh is of course correct. If you

simply to be able to filter or perform some

light weight reporting, my solution will work. 
for large volumes or more complex tracking then better to use the proper tooling but of course that is a major investment and maybe beyond the he scope of your project.