diff --git a/rattail_tempmon/server.py b/rattail_tempmon/server.py
index 64daa78..82851ac 100644
--- a/rattail_tempmon/server.py
+++ b/rattail_tempmon/server.py
@@ -30,7 +30,9 @@ import time
 import datetime
 import logging
 
+import six
 import humanize
+from sqlalchemy.exc import OperationalError
 
 from rattail.db import Session, api
 from rattail_tempmon.db import Session as TempmonSession, model as tempmon
@@ -53,13 +55,16 @@ class TempmonServerDaemon(Daemon):
         Keeps an eye on tempmon readings and sends alerts as needed.
         """
         self.extra_emails = self.config.getlist('rattail.tempmon', 'extra_emails', default=[])
+        delay = self.config.getint('rattail.tempmon', 'server.delay', default=60)
+        self.failed_checks = 0
+
         while True:
             self.check_readings()
-
-            # TODO: make this configurable
-            time.sleep(60)
+            time.sleep(delay)
 
     def check_readings(self):
+
+        # log.debug("checking readings")
         self.now = make_utc()
         session = TempmonSession()
 
@@ -69,11 +74,40 @@ class TempmonServerDaemon(Daemon):
                              .filter(tempmon.Client.archived == False)
             for client in clients:
                 self.check_readings_for_client(session, client)
-        except:
-            log.exception("Failed to check client probe readings (but will keep trying)")
+            session.flush()
+
+        except Exception as error:
+            log_error = True
+            self.failed_checks += 1
             session.rollback()
-        else:
+
+            # our goal here is to suppress logging when we see connection
+            # errors which are due to a simple postgres restart.  but if they
+            # keep coming then we'll go ahead and log them (sending email)
+            if isinstance(error, OperationalError):
+
+                # this first test works upon first DB restart, as well as the
+                # first time after DB stop.  but in the case of DB stop,
+                # subsequent errors will instead match the second test
+                if error.connection_invalidated or (
+                        'could not connect to server: Connection refused' in six.text_type(error)):
+
+                    # only suppress logging for 3 failures, after that we let them go
+                    # TODO: should make the max attempts configurable
+                    if self.failed_checks < 4:
+                        log_error = False
+                        log.debug("database connection failure #%s: %s",
+                                  self.failed_checks,
+                                  six.text_type(error))
+
+            # send error email unless we're suppressing it for now
+            if log_error:
+                log.exception("Failed to check client probe readings (but will keep trying)")
+
+        else: # checks were successful
+            self.failed_checks = 0
             session.commit()
+
         finally:
             session.close()