"Could not load suggestions" error when labeling documents in Custom Extractor

I’m trying to label freshly imported documents in Custom Extractor by clicking “Start Labeling” button. Normal behavior is for AI to pre-label detected fields. But this morning, I’m getting “Could not load suggestions” UI error.
Looking at the dev tools network tab I see an error message referencing pretrained-foundation-model-v1.0-2023-08-22 model. Where is this model coming from??? Yesterday the labeling process worked just fine.

“message”: “Processor 4c5e47bd6daa64f8 version pretrained-foundation-model-v1.0-2023-08-22 of type CUSTOM_EXTRACTION_PROCESSOR is no longer supported for sync processing. Please upgrade to a newer version: ``https://cloud.google.com/document-ai/docs/manage-processor-versions.”``,
“errorType”: “DATA_FETCHING_EXCEPTION”,
“path”: [
“processDocumentJspb2”
]

More on this bug.
Looks like detailed description for a derived signature field is causing this issue. If I clear the description for this field the Labeling Task works successfully.
Question to Google devs - why is the Labeling Task trying to use outdated pretrained-foundation-model-v1.0-2023-08-22 model when my processor’s default model is pretrained-foundation-model-v1.5-pro-2025-06-20?