diff --git a/test/.gitignore b/test/.gitignore
index 8f8929182..adbe0f597 100644
--- a/test/.gitignore
+++ b/test/.gitignore
@@ -3,9 +3,6 @@
 *.data
 *.log
 *.runlog
-_run_all_parallel.txt
-_run_all_serial.txt
-_run_all_parallel_rc1
 autocomplete.t
 color.t
 config.t
@@ -56,5 +53,3 @@ variant_xor.t
 view.t
 
 json_test
-
-run_all
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
index 288061ba3..4dd6f2f88 100644
--- a/test/CMakeLists.txt
+++ b/test/CMakeLists.txt
@@ -33,8 +33,6 @@ endif (CYGWIN)
 
 endif (${CMAKE_SOURCE_DIR} STREQUAL ${CMAKE_BINARY_DIR})
 
-configure_file (run_all.in run_all)
-
 add_custom_target (test ./run_all --verbose
                         DEPENDS ${test_SRCS} task_executable
                         WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/test)
diff --git a/test/README b/test/README
index be1b652f8..b8cc4e5eb 100644
--- a/test/README
+++ b/test/README
@@ -22,9 +22,10 @@ failing tests.
 
 Any TAP harness may be used.
 
-Note that adding the '--fast' option to ./run_all, the Python and C++ tests all
-run in parallel, alongside the Perl test that run serially. The result is a much
-quicker test run.
+Note that adding the '--serial' option to ./run_all, all tests are executed serially.
+The default runs Python and C++ tests in parallel, alongside the Perl tests
+that run serially (due to isolation limitations).
+Using '--serial' will make for a slower test run.
 
 
 Architecture
diff --git a/test/run_all b/test/run_all
new file mode 100755
index 000000000..abdf75aec
--- /dev/null
+++ b/test/run_all
@@ -0,0 +1,206 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+from __future__ import print_function
+import os
+import sys
+import glob
+import argparse
+import logging
+import time
+from multiprocessing import cpu_count
+from Queue import Queue, Empty
+from threading import Thread
+from subprocess import call, Popen, PIPE
+
+# Look for taskd in $PATH instead of task/src/
+os.environ["TASKD_USE_PATH"] = "1"
+
+TIMEOUT = .2
+
+
+def run_test(testqueue, outqueue):
+    while True:
+        try:
+            test = testqueue.get(block=True, timeout=TIMEOUT)
+        except Empty:
+            break
+
+        log.info("Running test %s", test)
+
+        try:
+            p = Popen(os.path.abspath(test), stdout=PIPE, stderr=PIPE,
+                      env=os.environ)
+            out, err = p.communicate()
+        except Exception as e:
+            log.exception(e)
+            # Premature end
+            break
+
+        output = ("# {0}\n".format(os.path.basename(test)), out, err)
+        log.debug("Collected output %s", output)
+        outqueue.put(output)
+
+        testqueue.task_done()
+
+
+class TestRunner(object):
+    def __init__(self):
+        self.threads = []
+        self.tap = open(cmd_args.tapfile, 'w')
+
+        self._parallelq = Queue()
+        self._serialq = Queue()
+        self._outputq = Queue()
+
+    def _find_tests(self):
+        for test in glob.glob("*.t") + glob.glob("*.t.exe"):
+            if os.access(test, os.X_OK):
+                # Executables only
+                if not cmd_args.serial:
+                    with open(test) as fh:
+                        if "/usr/bin/env python" in fh.readline():
+                            log.debug("Treating as parallel: %s", test)
+                            self._parallelq.put(test)
+                        else:
+                            log.debug("Treating as serial: %s", test)
+                            self._serialq.put(test)
+                else:
+                    log.debug("Treating %s as serial", test)
+                    self._serialq.put(test)
+            else:
+                log.debug("Ignored test %s as it is not executable", test)
+
+        log.info("Parallel tests: %s", self._parallelq.qsize())
+        log.info("Serial tests: %s", self._serialq.qsize())
+
+    def _prepare_threads(self):
+        # Serial thread
+        self.threads.append(
+            Thread(target=run_test, args=(self._serialq, self._outputq))
+        )
+        # Parallel threads
+        self.threads.extend([
+            Thread(target=run_test, args=(self._parallelq, self._outputq))
+            for i in range(cpu_count())
+        ])
+        log.info("Spawned %s threads to run tests", len(self.threads))
+
+    def _start_threads(self):
+        for thread in self.threads:
+            # Threads die when main thread dies
+            log.debug("Starting thread %s", thread)
+            thread.daemon = True
+            thread.start()
+
+    def _print_timestamp_to_tap(self):
+        now = time.time()
+        timestamp = "# {0} ==> {1}\n".format(
+            now,
+            time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(now)),
+        )
+
+        log.debug("Adding timestamp %s to TAP file", timestamp)
+        self.tap.write(timestamp)
+
+    def _get_remaining_tests(self):
+        return self._parallelq.qsize() + self._serialq.qsize()
+
+    def is_running(self):
+        for thread in self.threads:
+            if thread.is_alive():
+                return True
+
+        return False
+
+    def start(self):
+        self._find_tests()
+        self._prepare_threads()
+
+        self._print_timestamp_to_tap()
+
+        finished = 0
+        total = self._get_remaining_tests()
+
+        self._start_threads()
+
+        while self.is_running() or not self._outputq.empty():
+            try:
+                outputs = self._outputq.get(block=True, timeout=TIMEOUT)
+            except Empty:
+                continue
+
+            log.debug("Outputting to TAP: %s", outputs)
+
+            for output in outputs:
+                self.tap.write(output)
+
+                if cmd_args.verbose:
+                    sys.stdout.write(output)
+
+            self._outputq.task_done()
+            finished += 1
+
+            log.warning("Finished %s out of %s tests", finished, total)
+
+        self._print_timestamp_to_tap()
+
+        if not self._parallelq.empty() or not self._serialq.empty():
+            raise RuntimeError(
+                "Something went wrong, not all tests were ran. {0} "
+                "remaining.".format(self._get_remaining_tests()))
+
+    def show_report(self):
+        self.tap.flush()
+
+        log.debug("Calling 'problems --summary' for report")
+        call([os.path.abspath("problems"), "--summary", cmd_args.tapfile])
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(description="Run Taskwarrior tests")
+    parser.add_argument('--verbose', '-v', action="store_true",
+                        help="Also send TAP output to stdout")
+    parser.add_argument('--logging', '-l', action="count",
+                        help="Logging level. -lll is the highest level")
+    parser.add_argument('--serial', action="store_true",
+                        help="Do not run tests in parallel")
+    parser.add_argument('--tapfile', default="all.log",
+                        help="File to use for TAP output")
+    return parser.parse_args()
+
+
+def main():
+    runner = TestRunner()
+    runner.start()
+    if not cmd_args.verbose:
+        runner.show_report()
+
+
+if __name__ == "__main__":
+    cmd_args = parse_args()
+
+    if cmd_args.logging == 1:
+        level = logging.WARN
+    elif cmd_args.logging == 2:
+        level = logging.INFO
+    elif cmd_args.logging >= 3:
+        level = logging.DEBUG
+    else:
+        level = logging.ERROR
+
+    logging.basicConfig(
+        format="# %(asctime)s - %(levelname)s - %(message)s",
+        level=level,
+    )
+    log = logging.getLogger(__name__)
+
+    log.debug("Parsed commandline arguments: %s", cmd_args)
+
+    try:
+        main()
+    except Exception as e:
+        log.exception(e)
+        sys.exit(1)
+
+# vim: ai sts=4 et sw=4
diff --git a/test/run_all.in b/test/run_all.in
deleted file mode 100755
index e9d474aeb..000000000
--- a/test/run_all.in
+++ /dev/null
@@ -1,203 +0,0 @@
-#! /bin/sh
-
-# Look for taskd in $PATH instead of task/src/
-export TASKD_USE_PATH=1
-
-
-runlog_cleanup() {
-  if [ -f "_run_all_parallel.txt" ]; then
-    rm _run_all_parallel.txt
-  fi
-  if [ -f "_run_all_serial.txt" ]; then
-    rm _run_all_serial.txt
-  fi
-  if [ -f "_run_all_parallel_rc1" ]; then
-    rm _run_all_parallel_rc1
-  fi
-  for i in *.runlog; do
-    # Ugly hack. :)
-    if [ -f "$i" ]; then
-      rm *.runlog
-    fi
-    break
-  done
-}
-
-get_numprocs() {
-  numprocs=""
-  # Most Linux systems and OSX have getconf and _NPROCESSORS_ONLN.
-  if command -v getconf >/dev/null 2>&1; then
-    numprocs=$(getconf _NPROCESSORS_ONLN 2>/dev/null)
-  fi
-
-  # OpenBSD doesn't know _NPROCESSORS_ONLN, but it does have hw.ncpu
-  if [ "$numprocs" = "" ] && command -v sysctl >/dev/null 2>&1; then
-    numprocs=$(sysctl -n hw.ncpu 2>/dev/null)
-  fi
-
-  # If we still haven't found the number of CPU cores available, give up.
-  if [ "$numprocs" = "" ] || [ "$numprocs" -lt 1 ]; then
-    echo "Couldn't find number of CPU cores for parallelization. Assuming 2." 1>&2
-    numprocs=2
-  else
-    numprocs=$((numprocs+1))
-  fi
-
-  echo $numprocs
-}
-
-run_all_parallel() {
-  numprocs=$(get_numprocs)
-  cat _run_all_parallel.txt | xargs -n 1 -P $numprocs sh -c 'echo "#" $0 > $0.runlog; $0 >> $0.runlog 2>&1'
-  if [ $? -ne 0 ]; then
-    touch _run_all_parallel_rc1
-  fi
-  rm _run_all_parallel.txt
-}
-
-if [ ! -z "$1" ] && [ "$1" != "--verbose" ] && [ "$1" != "--fast" ];
-then
-  echo "Did you mean --fast or --verbose?"
-  exit 1
-fi
-
-if [ "$#" -gt 1 ];
-then
-  echo "Can only use arguments one at a time."
-  exit 1
-fi
-
-rc=0
-if [ x"$1" = x"--verbose" ];
-then
-  for i in ${TESTBLOB}
-  do
-    if [ -x "$i" ]; then
-      echo '#' $i
-      $i > test.log 2>&1
-      while read LINE
-      do
-        echo "$LINE"
-      done < test.log
-      if [ $? -ne 0 ]; then
-        rc=1
-      fi
-      rm test.log
-    else
-      echo "# Skipping $(basename $i) execute bit not set"
-    fi
-  done
-  exit $rc
-
-elif [ "$1" = "--fast" ]; then
-  # Useful for faster local testing, might not be portable. Use at own risk.
-  # Results in (almost) the exact same "all.log" as a normal run.
-  # Ordering is off, but could easily be adjusted to be the same.
-
-  date +"# %s ==> %a %b %d %H:%M:%S %Z %Y" > all.log
-
-  # Perl is used here to get the time in seconds
-  # because 'date +%s' isn't supported on Solaris.
-  STARTEPOCH=`perl -e 'print time'`
-
-  # Clean up after aborted runs
-  runlog_cleanup
-
-  for i in ${TESTBLOB}; do
-    if [ -x "$i" ]; then
-      # Only Python tests are guaranteed to run isolated.
-      if head -n 1 "$i" | grep -q '/usr/bin/env python'; then
-        echo $i >> _run_all_parallel.txt
-      else
-        echo $i >> _run_all_serial.txt
-      fi
-    else
-      echo "# Skipping $(basename $i) execute bit not set" >> all.log 2>&1
-    fi
-  done
-
-  run_all_parallel&
-
-  while read i; do
-    echo '#' $i >>all.log
-
-    $i >> all.log 2>&1
-    if [ $? -ne 0 ]; then
-      rc=1
-    fi
-  done < _run_all_serial.txt
-
-  while [ -f "_run_all_parallel.txt" ]; do
-    # Wait for the parallelized tests to finish running.
-    sleep 1  # sleep 0.1 is not portable.
-  done
-
-  if [ -f "_run_all_parallel_rc1" ]; then
-    rc=1
-  fi
-
-  cat *.runlog >> all.log
-
-  runlog_cleanup
-
-  date +"# %s ==> %a %b %d %H:%M:%S %Z %Y" >> all.log
-
-  ENDEPOCH=`perl -e 'print time'`
-  RUNTIME=`expr $ENDEPOCH - $STARTEPOCH`
-
-  printf "Pass:    %5d\n" `grep -c '^ok' all.log`
-  printf "Fail:    %5d\n" `grep -c '^not' all.log`
-  printf "Skipped: %5d\n" `grep -c '^skip' all.log`
-  printf "Runtime: %5d seconds\n" $RUNTIME
-  exit $rc
-
-else
-  date +"# %s ==> %a %b %d %H:%M:%S %Z %Y" > all.log
-
-  # Perl is used here to get the time in seconds
-  # because 'date +%s' isn't supported on Solaris.
-  STARTEPOCH=`perl -e 'print time'`
-
-  VRAMSTEG=`which vramsteg 2>/dev/null`
-  BAR=0
-  if [ -x "$VRAMSTEG" ]; then
-    BAR=1
-    COUNT=0
-    TOTAL=`ls ${TESTBLOB} | wc -l`
-    START=`$VRAMSTEG --now`
-  fi
-
-  for i in ${TESTBLOB}
-  do
-    if [ -x "$i" ]; then
-      echo '#' $i >>all.log
-
-      $i >> all.log 2>&1
-      if [ $? -ne 0 ]; then
-        rc=1
-      fi
-    else
-      echo "# Skipping $(basename $i) execute bit not set" >> all.log 2>&1
-    fi
-
-    if [ $BAR -eq 1 ]; then
-      $VRAMSTEG --label 'All tests' --min 0 --max $TOTAL --current $COUNT --percentage --start $START --estimate
-      COUNT=`expr $COUNT + 1`
-    fi
-  done
-
-  if [ $BAR -eq 1 ]; then
-    $VRAMSTEG --remove
-  fi
-
-  date +"# %s ==> %a %b %d %H:%M:%S %Z %Y" >> all.log
-
-  ENDEPOCH=`perl -e 'print time'`
-  RUNTIME=`expr $ENDEPOCH - $STARTEPOCH`
-
-  printf "Pass:    %5d\n" `grep -c '^ok' all.log`
-  printf "Fail:    %5d\n" `grep -c '^not' all.log`
-  printf "Skipped: %5d\n" `grep -c '^skip' all.log`
-  printf "Runtime: %5d seconds\n" $RUNTIME
-  exit $rc
-fi