It took about 2 weeks from reproducing the bug to understanding what the problem was, and then some more time before the bug in L was discovered.

The integrator back-ported the change from M - its only a few lines of code.

I researched mitigation (copying more data), but it was more trouble than it was worth. The problems were twofold: 1) increase in complexity of the state machine in the handler, 2) reliably reporting when the mitigation was happening, since it implied a bug we wanted to discover.