Hi Bob, I can't tell you what reports I want or what we'll do with the data until you've first collected the data for analysis. Thanks!
When you do finally figure out why you are in business, you may find that all the data you gathered is worthless. Sometimes it's really innocuous stuff like storing datetime values without enough resolution or accuracy. Only when you know how you plan to use the data can you ensure you are gathering it properly.
Teaching developers to collect all the data they can in hopes of being able to write arbitrary reports down the line is reckless behavior. Hyperscalers love pushing this ideology with inane bullshit like "Datalakes", so you can see where a conflict of interest may arise in the industry.
I've worked on a project (in a peripheral role, fortunately) where the entirety of the CIO's mandate was to "use AI", full stop. So the team was tasked with putting together a data warehouse that would collect random logs and app metrics, and then sprinkle AI fairy dust of their choice onto it to generate Insights(tm).
This kind of thing happens all. the time. in large companies where everybody involved wants either money or bullet points for their resume, and the actual outcome is nearly irrelevant.
I don't see how I could do it differently. I can only truly know which metrics are useful from experience with past incidents, but I can only gain this experience when the metrics were already collected during the past incident.