MDEV-11071, while technically unrelated, very badly affects current auto tests.
MDEV-13923 is still not in 10.3 tree, and it causes numerous failures, but at least it can be worked around by disabling geometry in the tests.
MDEV-12827 also shows up regularly and spoils results. No viable workaround.

Nov 14th

• MDEV-13201 appears a lot
• MDEV-13820 appears regularly
• MDEV-11071 still does too. Disabling temporary tables didn't help, it has other ways to fail.

• [Warning] InnoDB: Table test/t1 contains 24 user defined columns in InnoDB, but 23 columns in MariaDB. Please check INFORMATION_SCHEMA.INNODB_SYS_COLUMNS and http://dev.mysql.com/doc/refman/5.7/en/innodb-troubleshooting.html for how to resolve the issue.}} shows a lot in tests involving crash recovery, but that's expected

• 2017-11-15 10:44:21 0 [ERROR] InnoDB: Header page consists of zero bytes in datafile: ./test/#sql-62a8_10.ibd, Space ID:0, Flags: 0. Please refer to http://dev.mysql.com/doc/refman/5.7/en/innodb-troubleshooting-datadict.html for how to resolve the issue.

  2017-11-15 10:44:21 0 [ERROR] InnoDB: Datafile './test/#sql-62a8_10.ibd' is corrupted. Cannot determine the space ID from the first 64 pages.

• This I've only seen once so far:

  #5 0x00007fccad30a028 in __GI_abort () at abort.c:89

  #6 0x0000005591a321a2 in ut_dbg_assertion_failed (expr=0x0, file=0x5591f79828 "/home/travis/src/storage/innobase/log/log0recv.cc", line=1186) at /home/travis/src/storage/innobase/ut/ut0dbg.cc:61

  #7 0x00000055918a1c95 in recv_parse_or_apply_log_rec_body (type=MLOG_1BYTE, ptr=0x7fcc899d2318 "", end_ptr=0x7fcc899d231c "", space_id=144, page_no=0, apply=true, block=0x7fcc87b699b0, mtr=0x7fcc9497d550) at /home/travis/src/storage/innobase/log/log0recv.cc:1186

  #8 0x00000055918a4092 in recv_recover_page (just_read_in=true, block=0x7fcc87b699b0) at /home/travis/src/storage/innobase/log/log0recv.cc:1834

  #9 0x0000005591aa67fb in buf_page_io_complete (bpage=0x7fcc87b699b0, evict=false) at /home/travis/src/storage/innobase/buf/buf0buf.cc:6034

  #10 0x0000005591b3d073 in fil_aio_wait (segment=2) at /home/travis/src/storage/innobase/fil/fil0fil.cc:5359

  #11 0x00000055919cdbff in io_handler_thread (arg=0x55931a39f0 <n+16>) at /home/travis/src/storage/innobase/srv/srv0start.cc:340

  #12 0x00007fccadec1184 in start_thread (arg=0x7fcc9497e700) at pthread_create.c:312

  #13 0x00007fccad3cdffd in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111

The test set currently consists of 48 test runs – 12 combinations of grammar/data, each is run in 4 variants: normal, normal with server restarts, replication, replication with master restarts.
Example of results from one nightly set:

Last three positions are difficult to analyze, because attempts to repeat them usually end up with one of the already known MDEVs.

I think that the assertion failure in recv_parse_or_apply_log_rec_body() is very similar to MDEV-13101.

There appears to be a problem that can occur when PRIMARY KEY columns are updated and similar PRIMARY KEY are inserted.
So far, it has only been reported in the MDEV-12894 branch (bb-10.3-temporal), for the test

./mtr --mem --mysqld=--versioning-{force,hide=full,alter-history=keep} innodb.instant_alter,16k

See tempesta-tech#366 for some analysis.
A work-around for this bug is in tempesta-tech#367.
Edit: Here is a simplified version of the test case:

--source include/have_innodb.inc

CREATE TABLE t2

(id INT primary key, c1 VARCHAR(4000),

 p GEOMETRY NOT NULL DEFAULT ST_GeomFromText('LINESTRING(0 0,0 1,1 1)'))

ENGINE=InnoDB ROW_FORMAT=REDUNDANT;

INSERT INTO t2 (id,c1) VALUES (1, REPEAT('a', 4000)), (2, REPEAT('a', 4000));

ALTER TABLE t2 ADD COLUMN d1 INT;

BEGIN;

UPDATE t2 SET c1 = repeat(id, 4000);

ROLLBACK;

UPDATE t2 SET d1 = id;

DROP TABLE t2;

I converted this test case to reproduce the crash without using system versioning, and filed MDEV-14660.

10.3 and 10.4 went GA, so feature testing time has finished.