On February 1st, a scheduled batch job ran. It's a job that usually runs nightly with 127 batches and usually takes a couple minutes without issue. On this night it took much longer than this batch normally does (about 30 minutes) and had 10 extra batches. We have a log of the transactions this batch creates on external endpoint, with timestamps. What happened, per the endpoint transaction log, is the job paused for 10 minutes, repeated the prior 2 batches (resubmitted the transactions on the endpoint as if new, in close to exact same order minus a couple records), and then continued with the rest of the job's batches. 

I've checked with the managed app that runs the batch: the developer says it has no code that will pause or restart batches. It simply submits the data to the endpoint with this job, batching through records. (This is Chargent scheduled recurring transactions)

I've checked with the endpoint developers and they see nothing odd on their end, just the log that the transactions were submitted to them in this pattern-- they stopped coming in for 10 minutes, repeated the last couple batches worth as if newly submitted in the same order, then continued to complete the batches as per normal.

The only other place we have to look is that the batch engine did something strange. The fact that it appears to have paused, reran some batches and then continues processing lends me to believe that something went wrong in the underlying batch process itself, since it runs every other night just fine. 

My access to apex job logs is only 7 days back of course, but I have a screen shot of the jobs list from that time-frame displaying this odd behaviour:  https://drive.google.com/file/d/1y6Dy2TFedF8e_AwX9qaeNLGt8jDfPW2z/view?usp=sharing  (Ignore the batch of 190, I dropped batch size just in case. Up until that point it was 127 a night. )

I have no course to submit this to support because this nonprofit client doesn't have the right support level, so thanks for any ideas!