Answer
In a case that reached this same dead end after similarly exhausting targeted troubleshooting (and working with IFS Support without finding a definitive root cause), the resolution was a full redeployment of the IFS Application from the IFS Build onto a brand new server — after which the instability didn't return. This doesn't diagnose what was actually wrong (likely some form of accumulated environment-level drift or corruption that resisted targeted elimination), but it's worth keeping in mind as a genuine last-resort option: if a recurring, load-correlated instability survives disabling every individually suspected component without resolving, redeploying onto a clean server from a known-good Build is a real (if blunt) way some environments have recovered when the actual root cause couldn't be pinned down.