Help families in GazaDonate to the IRC to provide humanitarian aid and resettlement resources to families in Gaza
JT's Space
Articles
About
Journal
RaW
GitHub
Articles
About
Journal
RaW
GitHub

Wednesday, July 8, 2026

JT Houk
Published July 8, 2026
code-reviewinvestigations

Projects

The QBO backfill dry runs all completed cleanly and the performance looked good, which was encouraging. The first real execute run, however, surfaced a shard isolation problem that had been lurking under the surface. This is the kind of thing dry runs often miss because they are not doing real writes, so it is not a failure of the testing approach so much as a reminder that some bugs only show themselves when the stakes get real. I wrote and opened PR #23304 to fix it by clearing the stale identity on the shared QBO integration singleton, which is the LRX-41775 issue I filed on Friday. Getting that singleton into a clean state on each run is the right fix and should prevent identity bleed from quietly corrupting future syncs.

I also reviewed Vicky's PR #16765 on the monolith, covering LRX-38507 to update the return and exchange flow to use state and attributes. This is the same pattern we landed in the EditSale and fulfil sale work earlier, and it is good to see it propagating into the return and exchange path. Consistency across the full sale lifecycle is the goal, and we are getting closer to that.

External Requests

I took a look at PR #7136 in webregister covering LRX-41845, which addresses a stale onAccountBalance in customer balances. Stale balance values are one of those issues that feel minor in isolation but erode trust fast when a cashier sees numbers that do not match reality at the point of sale, so getting this right matters.

Skip to main content