No error handling of "no alive nodes"

Nextcloud 14.0.8
fulltextsearch 1.1.1
fts-elasticsearch 1.0.3
fts-files 1.1.3

While running occ fulltextsearch:index, sometimes the Elasticsearch platform crashes for different reasons (unable to parse content, oom, whatever -- not relevant here). From this moment, all files being processed get marked as 'processing done, but an error occured'. The database (table oc_fulltextserach_indexes) stores the 'no alive nodes' error for each and every file. This prevents reindexing those files until the index is discarded -- not a sensible solution given 10M files -- reindexing could easily take a month or more.

I understand that content-related errors make sense to be registered that way, just to avoid reindexing those files over and over. But why is 'no nodes available' registered that way?

Start with the occ fulltextsearch:index path and inspect how Elasticsearch's "no alive nodes" error is stored in oc_fulltextserach_indexes. Confirm that content-related errors remain recorded while a platform outage leaves affected files eligible for reindexing without discarding the whole index.