On a very large corpus, initial scan operations can take a long time, and build up quite a large debuginfod.sqlite-wal file. If the server is filesystem space constrained, this .sqlite + .sqlite-wal combined file may get too large to fit. Suggestion: debuginfod should inject a periodic pragma wal_checkpoint(truncate); into the traverse loop, as the default autocheckpoint stuff does not appear to operate reliably.
commit d16628bb7dc39a020e555e16bf56374fdac84e42 Author: Frank Ch. Eigler <fche@redhat.com> Date: Mon May 8 11:05:48 2023 -0400 PR30316: debuginfod wal checkpointing