N1QL index goes from over 30 GB data size down to 0 bytes

We are running version 4.1.1-5914 Enterprise Edition (build-5914).

To improve performance and provide high availability, we have identical N1QL indexes configured on 2 nodes in the cluster. For the last couple weeks we have been populating the database with documents. The indexes on both nodes were fine until I noticed today that an index on one node was showing (from the Couchbase console) the following:
8 KB disk size
0 B data size
11 K total items remaining
95.5 M total indexed items
And the identical “good” index on the other node shows
39.6 GB disk size
34.6 GB data size
15.5 K total items remaining
95.5 M total indexed items

Below are the index files on the bad node:
[root@lin153-virt usr_deleted_startTime___c7a6fb7f133_4171947888624910125_0.index]# ls -lart
total 156767100
drwxr-x— 7 couchbase couchbase 4096 Aug 8 11:29 …
drwxrwx— 2 couchbase couchbase 4096 Aug 23 16:45 .
-rw-rw---- 1 couchbase couchbase 8192 Aug 23 21:20 data.fdb.148
-rw-rw---- 1 couchbase couchbase 160824733696 Aug 25 17:25 data.fdb.149

Notice the data.fdb.148 file is 8Kb (which is disk size displayed on the console), and the newer file data.fdb.149 is ~150GB

Have you seen this behavior before, and is there a way to recover from it?
In addition, I have saved off the logs from “the bad node” for the last couple of days. If you want the log files (many of them, and fairly large) just let me know how to get them to you?