Performance regression

[casey]

I triggered a database rebuild on ordinals.com last night, using commit 1ad94af. It should have been done by now, but it's still at block height ~265000. Indexing is slow enough that I suspect it might not finish, or take an enormously long time.

The last deployed version was 0.3.0, and looking through the commit history since then, there have been a few suspicious commits:

• Upgrade to redb 0.11.0
• Track inscriptions at offset and vout other than first
• Track inscription locations correctly

[casey]

I'm rebuilding the database with redb 0.10.0, to narrow it down to either an issue with redb, or an issue with one of our changes. #1167

[casey]

Not done yet, but it looks like it's still slow with redb 0.10.0. I'm thinking that it's the other two PRs. In particular, they added writes mid-update, instead of only flushing at the end of the update.

If it's definitely not redb, then I should modify the updater to only flush at the end of each update.