Number of node_access records peaked and then went LOWER during drush run

Problem/Motivation

While running drush node-access-rebuild-progressive

Occasional checks of the number of node access records created,

SELECT COUNT(*) FROM node_access WHERE realm = 'hw_provider_program';

showed it rising to

98221

… and then it started going down, which i am struggling to see how this could be possible no matter how messed up my code is.

And it is not like it reset to zero and went back up, it is going down at the same rate it went up, a thousand or so a minute.

It did ultimately end without error, with 84382 records for the relevant realm in node_access.

Does anybody have any idea what might have happened?

Steps to reproduce

I doubt this can be reproduced (and really don't know if i hope it can be or not!), but it did happen on the run after a run that had an error (the computer went to sleep for the night while running it, so i am fairly certain that was the cause of the error):

  The process "/var/www/html/vendor/bin/drush php-eval '_drush_node_access_rebuild_progressive_process(512823);' --format=json --uri=https://app-housingworks-net.ddev.site" exceeded the timeout of 14400 sec  
  onds.  

So maybe it is related to / could be fixed by the work in 🐛 "drush node-access-rebuild-progressive" should not rebuild access grants if not needed Needs work

Proposed resolution

Remaining tasks

User interface changes

API changes

Data model changes