Internal Server Error after updating custom Language Model for Watson's speech2text service - speech-to-text

Question

We've had success using a custom Language Model for Watson's speech2text service. However, recently, we decided to extend the model (extended vocabulary, utterances, etc.). To that end, we reset the model:
curl -X POST -u ${credentials}
"https://stream.watsonplatform.net/speech-to-text/api/v1/customizations/${customization_id}/reset"
added our extended corpus:
curl -X POST -u ${credentials} \
--data-binary #corpus.txt \
"https://stream.watsonplatform.net/speech-to-text/api/v1/customizations/${customization_id}/corpora/corpus_test01"
verified the process completed, and re-trained the model:
curl -X POST -u ${cred} \
"https://stream.watsonplatform.net/speech-to-text/api/v1/customizations/${customization_id}/train"
Once the process completed ("status": "available"), we were initially able to get responses, and the recognition quality did improve. However, within a few minutes, we started getting {"error": "Internal Server Error"}. Can't reset or query. The base model works fine though. Is our only option to create a new model (we would prefer not to)?
Any help would be appreciated.
Revised: don't seem to be able to create a new LM either:
curl -X POST -u ${cred} \
--header "Content-Type: application/json" \
--data "{\"name\": \"Test model\", \
\"base_model_name\": \"en-US_BroadbandModel\", \
\"description\": \"language model 02\"}" \
"https://stream.watsonplatform.net/speech-to-text/api/v1/customizations"
returns:
{
"code": 500,
"code_description": "Internal Server Error",
"error": "Internal Server Error"
}
But base model still works fine ...
New update: LM came back several hours later, out of a blue, and kept working fine through the day. However, this evening became intermittent, returning "Internal Server Error" with increasing frequency. Not very usable at this point.
-rg