Not a rocket scientist; just an electrical engineer with plenty of failure analysis experience. Here is generally how I would translate my process to this case.
As a sibling mentioned, there will be lots of telemetry about what was going on with various systems during the incident. That data may indicate a few specific points to start, but it will mostly be useful for putting other facts into context. In parallel would be an effort to comb over the pad, recover as much as possible of what remains of the rocket, and try to identify what parts are what. There may be some obvious forensic evidence of parts to concentrate on, but in all likelihood the analysis will be guided by a combination of analyzing the telemetry and the damage patterns on the remaining pieces. Hopefully the combination will narrow the probable causes down enough to concentrate analysis on a couple-three theories.
Once plausible and probable theories are identified, teams will start to drill down into what event chains could cause them and look for evidence to confirm or reject those. Some of that will involve destructive analysis of the wreckage; you can glean a surprising amount of information from high-magnification images, x-rays, and sections of what otherwise looks like a twisted hunk of scrap metal. Some will involve modeling; no doubt the engineers who designed the engines have some sort of model they used to test it, and the parameters on that model can be varied to create various out-of-spec conditions. Some will be design analysis, to see if some previously-unseen corner case could have caused an otherwise-conformant system to fail catastrophically. Test data on the involved systems will be analyzed to look for any anomalies that might have been passed off at the time but that might be significant on hindsight.
Assuming possibilities still exist and are not sufficiently firm, or even just to be extra thorough, a few possibilities may be tested in practical (and possibly destructive) tests. In other words, try to blow up (or simulate blowing up) another engine in a controlled manner. That may or may not be possible, depending on what the root cause is; a parts tolerance issue, for example, would be practically impossible to recreate unless it were possible to manually change affected parts to match the suspected tolerances that caused the failure.
I'm sure an actual rocket scientist will come along and provide more detail. My work was limited to exploding caps and FETs.
As a sibling mentioned, there will be lots of telemetry about what was going on with various systems during the incident. That data may indicate a few specific points to start, but it will mostly be useful for putting other facts into context. In parallel would be an effort to comb over the pad, recover as much as possible of what remains of the rocket, and try to identify what parts are what. There may be some obvious forensic evidence of parts to concentrate on, but in all likelihood the analysis will be guided by a combination of analyzing the telemetry and the damage patterns on the remaining pieces. Hopefully the combination will narrow the probable causes down enough to concentrate analysis on a couple-three theories.
Once plausible and probable theories are identified, teams will start to drill down into what event chains could cause them and look for evidence to confirm or reject those. Some of that will involve destructive analysis of the wreckage; you can glean a surprising amount of information from high-magnification images, x-rays, and sections of what otherwise looks like a twisted hunk of scrap metal. Some will involve modeling; no doubt the engineers who designed the engines have some sort of model they used to test it, and the parameters on that model can be varied to create various out-of-spec conditions. Some will be design analysis, to see if some previously-unseen corner case could have caused an otherwise-conformant system to fail catastrophically. Test data on the involved systems will be analyzed to look for any anomalies that might have been passed off at the time but that might be significant on hindsight.
Assuming possibilities still exist and are not sufficiently firm, or even just to be extra thorough, a few possibilities may be tested in practical (and possibly destructive) tests. In other words, try to blow up (or simulate blowing up) another engine in a controlled manner. That may or may not be possible, depending on what the root cause is; a parts tolerance issue, for example, would be practically impossible to recreate unless it were possible to manually change affected parts to match the suspected tolerances that caused the failure.
I'm sure an actual rocket scientist will come along and provide more detail. My work was limited to exploding caps and FETs.