When @paraseba wrote this code, he was thinking of a "nice to have". But this feature has become critical to several ideas on server-side Icechunk usage, like diff detection and synchronization. So, now we need to make sure it's actually well implemented.
First step, I think, would be a careful review of the code, thinking carefully about all the places where the process could be interrupted or preempted by some other process.
We need to have very high levels of confidence that the ops_log is always correct in all scenarios.
When @paraseba wrote this code, he was thinking of a "nice to have". But this feature has become critical to several ideas on server-side Icechunk usage, like diff detection and synchronization. So, now we need to make sure it's actually well implemented.
First step, I think, would be a careful review of the code, thinking carefully about all the places where the process could be interrupted or preempted by some other process.
We need to have very high levels of confidence that the
ops_logis always correct in all scenarios.