Fix missing label retry state leaking across ontology processing

[alexskr]

Addresses

• submission_missing_labels leaks request-scoped state after retry/failure and contaminates later ontology processing #271

[mdorf]

One minor issue: silent success on page retrieval failure

Suppose paging.page(page, size).all at line 110 raises a StandardError (e.g., a SPARQL timeout, network. error, or triplestore issue).

Step 1 — The error is caught and swallowed (lines 109-117):

begin
  page_classes = paging.page(page, size).all   # raises StandardError
  total_pages = page_classes.total_pages
  page_len = page_classes.length
rescue StandardError => e
  page_classes = []          # line 114
  logger.error(...)          # line 115 — logged, but execution continues
  logger.flush
end

After the rescue: page_classes = [], total_pages = 0 (from line 107 initialization).

Step 2 — The retry block is skipped (line 120):

if total_pages > 0 && page_classes.empty? && ...

Since total_pages = 0, this entire block (lines 120-147) is skipped. The retry logic never fires — the error is treated as "there are simply no pages."

Step 3 — The loop breaks (lines 149-156):

if page_classes.empty?
  if total_pages > 0
    ...
  else
    logger.info("Ontology #{acr} contains #{total_pages} pages...")  # line 153
  end
  break   # line 155
end

The loop exits cleanly with a log message that says the ontology contains 0 pages — as if this is a normal, expected condition rather than an error.

Step 4 — Success status is set (lines 187-191):

# set the status on actions that have completed successfully
callbacks.each do |_, callback|
  if callback[:status]
    @submission.add_submission_status(callback[:status])   # line 189 — marks RDF_LABELS as SUCCESS
    @submission.save
  end
end

The submission is marked with the RDF_LABELS success status even though no labels were actually processed — the page fetch failed with an exception.

Step 5 — handle_missing_labels rescue is never triggered (lines 33-38):

Since loop_classes returned normally (no exception propagated), the rescue Exception => e at line 33 never fires, so the error status is never set.

Suggested fix

Re-raise the exception after logging on the initial page fetch, so it reaches the error status path in handle_missing_labels. The retry-loop rescue (lines 128-138) can remain as-is since the retry block already handles exhausted retries by raising at line 145.

# lines 109-117: re-raise on initial fetch failure
begin
  page_classes = paging.page(page, size).all
  total_pages = page_classes.total_pages
  page_len = page_classes.length
rescue StandardError => e
  page_classes = []
  logger.error("Error retrieving page #{page} for #{acr}: #{e.class}: #{e.message}")
  logger.flush
  raise e  # let handle_missing_labels rescue set the error status
end