VRM backlog only sends with smaller batch

I had a problem with VRM on my Cerbo GX. The backlog kept growing and was not uploading.

The error was:

Read timed out (read timeout=5)

Internet and HTTPS connection to the Victron server were working normally.

I found that vrmlogger was trying to send the backlog in batches of 100 records. I changed:

limit=100

to:

limit=5

After restarting vrmlogger, it immediately started sending the backlog. The queue went from over 1000 records to zero and VRM is working normally again.

I restored the HTTP timeout back to the original 5 seconds.

Has anyone else seen this problem? Could the batch size of 100 be too large on some Internet connections?