The PALI materialization concensus frontier is calculated globally, rather than per segment, which leads to the following kind of outage:

• For segment 3: cells A and B are fully_materialized but C is not
• For segment 5: A and C are fully_materialized, and B is not.

Individually both segments 3 and 5 are "healthy" (2/3 completed), but globally PALI will perceive a mat outage where cell C is pinned in segment 3 and B is pinned at 5.

In order to solve this problem we could take the concensus frontier for each segment individually (i.e. both 3 and 5 are concensus-fully-materialized so the global log should be).

The immediate problem with the above is to do with page read routing. In thie above situation you want to route segment 3 reads around C and segment 5 reads around B but the PSR doesn't have access to segment information or LSN boundary information.