Although i did not yet notice any problems when not resetting the ORDERLABEL for pages (as done in this PR), we can also extend the original logic. The root cause of the linked issue is, that the necessary logic for setting the LABEL and ORDERLABEL (which are set to NULL in the beginning) is not called for hidden metatdata.

To solve the issue at hand we could for hiddenMetadata call the setters directly to not lose the ORDERLABELs and therefor our pagination. We would by this also preserve hidden division LABEL's, which might be lost.

private void processHiddenMetadata() {
        for (Metadata metadatum : hiddenMetadata) {
            if (metadatum instanceof MetadataEntry) {
                MetadataEntry metadataEntry = (MetadataEntry) metadatum;
                String key = metadataEntry.getKey().toUpperCase();
                String value = metadataEntry.getValue();
                switch (key) {
                    case "LABEL":
                        division.setLabel(value);
                        break;
                    case "ORDERLABEL":
                        division.setOrderlabel(value);
                        break;
                    case "CONTENTIDS":
                        division.getContentIds().add(URI.create(value));
                        break;
                }
            }
        }
        metadata.addAll(hiddenMetadata);
    }

A more general remark: The necessary setters for the divisions of (non-hidden) metadata elements in the UI are called as a side effect of the (IMHO very complicated) BiConsumer logic, which does some transformations to the values from the tree, but mainly defines the setters for different division values, which are then applied. There may be a specific reason for using functional interfaces here, but I wonder if there might be opportunities to simplify the logic for better readability and maintainability.

@BartChris Please rebase against current master to avoid failing builds caused by broken Selenium tests!