Just to report that I got hit with the same issue on another node. I looked at the persistent queue directory and saw it had three page files - page.105, page.125 and page.126. It also had all the checkpoint files from 105 to 125 and `head. Removed all files from this directory and logstash started correctly.
I have another open question about "counter incremented with a negative value" that freezes pipeline workers which I suspect is to blame for "corrupted" persistent queues.