LOCKLOSS guardian stopped again by NDS connection error for the lockloss evnet at 05:37:27 JST on May 7th (20:37:27 UTC on May 6th).
I resumed it at 14:43 JST and reproduced missed lockloss events.
There was 14 lockloss events on May 6th UTC (Fig.1) and 9 events on May 7th (Fig.2).

NDS overload is more likely to occur.
Has a number of processes accessing NDS increased recently?