Reva/ocdav seems to have memory leak when doing webdav file transfers.

[redblom]

In experimenting with file transfers initiated by rclone/webdav I noticed that Reva runs out of memory very quickly with subsequent file transfers. The memory used for each transfer keeps adding up to the previously used memory until Reva runs out of memory, and then Reva is killed.
Here's the final part of the log until Reva is killed: mem-issue-reva-webdav-file-transfer.txt

It's the receiving Reva instance that takes up all the memory. In the experiments both source and destination Reva instances are running in docker containers as well as rclone.
Here's the result (docker stats) from a number of subsequent 600MB file transfers (4GB memory available for docker). Each subsequent transfer adds to the already used memory and in between transfers the memory does not gets freed up:

CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
2fd9dee452e9        reva-2              0.08%               1.777GiB / 3.848GiB   46.17%              615MB / 617kB       0B / 0B             10
2fd9dee452e9        reva-2              0.00%               2.668GiB / 3.848GiB   69.33%              1.23GB / 1MB        0B / 0B             10
2fd9dee452e9        reva-2              0.00%               3.225GiB / 3.848GiB   83.81%              1.84GB / 1.49MB     0B / 0B             11
2fd9dee452e9        reva-2              0.00%               3.434GiB / 3.848GiB   89.23%              2.46GB / 2.02MB     0B / 0B             12
2fd9dee452e9        reva-2              0.25%               3.547GiB / 3.848GiB   92.18%              2.56GB / 2.05MB     0B / 0B             12
 ... Reva KILLED

I am not experienced in the underlying implementation (TUS ?). Anyone?