diff --git a/Lib/_pydecimal.py b/Lib/_pydecimal.py
index 4b7bfd07887bd3..1e9434f323214e 100644
--- a/Lib/_pydecimal.py
+++ b/Lib/_pydecimal.py
@@ -62,6 +62,7 @@
 import math as _math
 import numbers as _numbers
 import sys
+from contextlib import contextmanager as _contextmanager
 
 try:
     from collections import namedtuple as _namedtuple
@@ -1499,8 +1500,9 @@ def __divmod__(self, other, context=None):
                 ans = context._raise_error(DivisionUndefined, 'divmod(0, 0)')
                 return ans, ans
             else:
-                return (context._raise_error(DivisionByZero, 'x // 0', sign),
-                        context._raise_error(InvalidOperation, 'x % 0'))
+                with context._defer_traps():
+                    return (context._raise_error(DivisionByZero, 'x // 0', sign),
+                            context._raise_error(InvalidOperation, 'x % 0'))
 
         quotient, remainder = self._divide(other, context)
         remainder = remainder._fix(context)
@@ -1746,9 +1748,10 @@ def _fix(self, context):
         exp_min = len(self._int) + self._exp - context.prec
         if exp_min > Etop:
             # overflow: exp_min > Etop iff self.adjusted() > Emax
-            ans = context._raise_error(Overflow, 'above Emax', self._sign)
-            context._raise_error(Inexact)
-            context._raise_error(Rounded)
+            with context._defer_traps():
+                ans = context._raise_error(Overflow, 'above Emax', self._sign)
+                context._raise_error(Inexact)
+                context._raise_error(Rounded)
             return ans
 
         self_is_subnormal = exp_min < Etiny
@@ -1770,34 +1773,36 @@ def _fix(self, context):
                     coeff = coeff[:-1]
                     exp_min += 1
 
-            # check whether the rounding pushed the exponent out of range
-            if exp_min > Etop:
-                ans = context._raise_error(Overflow, 'above Emax', self._sign)
-            else:
-                ans = _dec_from_triple(self._sign, coeff, exp_min)
+            # check if rounding pushed the exponent out of range
+            with context._defer_traps():
+                if exp_min > Etop:
+                    ans = context._raise_error(Overflow, 'above Emax', self._sign)
+                else:
+                    ans = _dec_from_triple(self._sign, coeff, exp_min)
+
+                # raise the appropriate signals, taking care to respect
+                # the precedence described in the specification
+                if changed and self_is_subnormal:
+                    context._raise_error(Underflow)
+                if self_is_subnormal:
+                    context._raise_error(Subnormal)
+                if changed:
+                    context._raise_error(Inexact)
+                context._raise_error(Rounded)
+                if not ans:
+                    # raise Clamped on underflow to 0
+                    context._raise_error(Clamped)
+            return ans
 
-            # raise the appropriate signals, taking care to respect
-            # the precedence described in the specification
-            if changed and self_is_subnormal:
-                context._raise_error(Underflow)
+        with context._defer_traps():
             if self_is_subnormal:
                 context._raise_error(Subnormal)
-            if changed:
-                context._raise_error(Inexact)
-            context._raise_error(Rounded)
-            if not ans:
-                # raise Clamped on underflow to 0
-                context._raise_error(Clamped)
-            return ans
-
-        if self_is_subnormal:
-            context._raise_error(Subnormal)
 
-        # fold down if clamp == 1 and self has too few digits
-        if context.clamp == 1 and self._exp > Etop:
-            context._raise_error(Clamped)
-            self_padded = self._int + '0'*(self._exp - Etop)
-            return _dec_from_triple(self._sign, self_padded, Etop)
+            # fold down if clamp == 1 and self has too few digits
+            if context.clamp == 1 and self._exp > Etop:
+                context._raise_error(Clamped)
+                self_padded = self._int + '0'*(self._exp - Etop)
+                return _dec_from_triple(self._sign, self_padded, Etop)
 
         # here self was representable to begin with; return unchanged
         return Decimal(self)
@@ -2445,8 +2450,9 @@ def __pow__(self, other, modulo=None, context=None):
                     exp = 1-context.prec
                     context._raise_error(Rounded)
             else:
-                context._raise_error(Inexact)
-                context._raise_error(Rounded)
+                with context._defer_traps():
+                    context._raise_error(Inexact)
+                    context._raise_error(Rounded)
                 exp = 1-context.prec
 
             return _dec_from_triple(result_sign, '1'+'0'*-exp, exp)
@@ -2554,11 +2560,12 @@ def __pow__(self, other, modulo=None, context=None):
             # Inexact, Rounded, Clamped.  Overflow needs the correct
             # arguments.  Note that the order of the exceptions is
             # important here.
-            if newcontext.flags[Overflow]:
-                context._raise_error(Overflow, 'above Emax', ans._sign)
-            for exception in Underflow, Subnormal, Inexact, Rounded, Clamped:
-                if newcontext.flags[exception]:
-                    context._raise_error(exception)
+            with context._defer_traps():
+                if newcontext.flags[Overflow]:
+                    context._raise_error(Overflow, 'above Emax', ans._sign)
+                for exception in Underflow, Subnormal, Inexact, Rounded, Clamped:
+                    if newcontext.flags[exception]:
+                        context._raise_error(exception)
 
         else:
             ans = ans._fix(context)
@@ -2645,17 +2652,18 @@ def quantize(self, exp, rounding=None, context=None):
             return context._raise_error(InvalidOperation,
                                         'quantize result has too many digits for current context')
 
-        # raise appropriate flags
-        if ans and ans.adjusted() < context.Emin:
-            context._raise_error(Subnormal)
-        if ans._exp > self._exp:
-            if ans != self:
-                context._raise_error(Inexact)
-            context._raise_error(Rounded)
+        with context._defer_traps():
+            # raise appropriate flags
+            if ans and ans.adjusted() < context.Emin:
+                context._raise_error(Subnormal)
+            if ans._exp > self._exp:
+                if ans != self:
+                    context._raise_error(Inexact)
+                context._raise_error(Rounded)
 
-        # call to fix takes care of any necessary folddown, and
-        # signals Clamped if necessary
-        ans = ans._fix(context)
+            # call to fix takes care of any necessary folddown, and
+            # signals Clamped if necessary
+            ans = ans._fix(context)
         return ans
 
     def same_quantum(self, other, context=None):
@@ -2754,9 +2762,10 @@ def to_integral_exact(self, rounding=None, context=None):
         if rounding is None:
             rounding = context.rounding
         ans = self._rescale(0, rounding)
-        if ans != self:
-            context._raise_error(Inexact)
-        context._raise_error(Rounded)
+        with context._defer_traps():
+            if ans != self:
+                context._raise_error(Inexact)
+            context._raise_error(Rounded)
         return ans
 
     def to_integral_value(self, rounding=None, context=None):
@@ -3646,21 +3655,22 @@ def next_toward(self, other, context=None):
             ans = self.next_minus(context)
 
         # decide which flags to raise using value of ans
-        if ans._isinfinity():
-            context._raise_error(Overflow,
-                                 'Infinite result from next_toward',
-                                 ans._sign)
-            context._raise_error(Inexact)
-            context._raise_error(Rounded)
-        elif ans.adjusted() < context.Emin:
-            context._raise_error(Underflow)
-            context._raise_error(Subnormal)
-            context._raise_error(Inexact)
-            context._raise_error(Rounded)
-            # if precision == 1 then we don't raise Clamped for a
-            # result 0E-Etiny.
-            if not ans:
-                context._raise_error(Clamped)
+        with context._defer_traps():
+            if ans._isinfinity():
+                context._raise_error(Overflow,
+                                     'Infinite result from next_toward',
+                                     ans._sign)
+                context._raise_error(Inexact)
+                context._raise_error(Rounded)
+            elif ans.adjusted() < context.Emin:
+                context._raise_error(Underflow)
+                context._raise_error(Subnormal)
+                context._raise_error(Inexact)
+                context._raise_error(Rounded)
+                # if precision == 1 then we don't raise Clamped for a
+                # result 0E-Etiny.
+                if not ans:
+                    context._raise_error(Clamped)
 
         return ans
 
@@ -3988,6 +3998,11 @@ def __init__(self, prec=None, rounding=None, Emin=None, Emax=None,
         else:
             self._ignored_flags = _ignored_flags
 
+        # _defer_traps state: nesting level and the deferred exception.
+        # Transient: not copied or pickled.
+        self._deferred_traps = 0
+        self._deferred_error = None
+
         if traps is None:
             self.traps = dc.traps.copy()
         elif not isinstance(traps, dict):
@@ -4046,7 +4061,7 @@ def __setattr__(self, name, value):
             return object.__setattr__(self, name, value)
         elif name == 'flags' or name == 'traps':
             return self._set_signal_dict(name, value)
-        elif name == '_ignored_flags':
+        elif name in {'_ignored_flags', '_deferred_traps', '_deferred_error'}:
             return object.__setattr__(self, name, value)
         else:
             raise AttributeError(
@@ -4134,10 +4149,42 @@ def _raise_error(self, condition, explanation = None, *args):
             # The errors define how to handle themselves.
             return condition().handle(self, *args)
 
+        if self._deferred_traps:
+            # Inside a _defer_traps block: record the first trapped
+            # exception instead of raising it, and return the default
+            # response as if the signal had not been trapped.
+            if self._deferred_error is None:
+                self._deferred_error = error(explanation)
+            return condition().handle(self, *args)
+
         # Errors should only be risked on copies of the context
         # self._ignored_flags = []
         raise error(explanation)
 
+    @_contextmanager
+    def _defer_traps(self):
+        """Defer trapped signals until the end of the with block.
+
+        _raise_error() sets the flag of every signal in the block and
+        raises the first trapped exception on exit, so that a
+        multi-signal operation sets the same flags as _decimal.
+        """
+        self._deferred_traps += 1
+        try:
+            yield
+        except BaseException:
+            # The block is aborted, so the deferred error is
+            # discarded: the propagating exception wins, and the
+            # context does not retain an exception object.
+            self._deferred_error = None
+            raise
+        finally:
+            self._deferred_traps -= 1
+        if self._deferred_traps == 0 and self._deferred_error is not None:
+            error = self._deferred_error
+            self._deferred_error = None
+            raise error
+
     def _ignore_all_flags(self):
         """Ignore all flags, if they are raised"""
         return self._ignore_flags(*_signals)
diff --git a/Lib/test/test_decimal.py b/Lib/test/test_decimal.py
index 424d3f577e9f87..64dca8e2a19f8b 100644
--- a/Lib/test/test_decimal.py
+++ b/Lib/test/test_decimal.py
@@ -4014,6 +4014,59 @@ def raise_error(context, flag):
                                   "operation raises different flags depending on flags set: " +
                                   "expected %s, got %s" % (expected_flags, new_flags))
 
+    def test_trapped_signal_sets_companion_flags(self):
+        # If an operation signals several conditions and one of them is
+        # trapped, the exception is raised, but the flags of all the
+        # other conditions encountered must be set as well.
+        Decimal = self.decimal.Decimal
+        Context = self.decimal.Context
+        Clamped = self.decimal.Clamped
+        DivisionByZero = self.decimal.DivisionByZero
+        Inexact = self.decimal.Inexact
+        InvalidOperation = self.decimal.InvalidOperation
+        Overflow = self.decimal.Overflow
+        Rounded = self.decimal.Rounded
+        Subnormal = self.decimal.Subnormal
+        Underflow = self.decimal.Underflow
+
+        def check(trap, operation, expected_flags):
+            context = Context(prec=9, Emin=-999, Emax=999, traps=[])
+            context.traps[trap] = 1
+            with self.assertRaises(trap):
+                operation(context)
+            flags = {signal for signal, v in context.flags.items() if v}
+            self.assertEqual(flags, expected_flags)
+
+        # overflow from _fix: Inexact and Rounded are also signaled
+        check(Overflow,
+              lambda c: c.multiply(Decimal('1E+800'), Decimal('1E+800')),
+              {Inexact, Overflow, Rounded})
+
+        # 1/3 is rounded: Rounded accompanies the trapped Inexact
+        check(Inexact,
+              lambda c: c.divide(Decimal(1), Decimal(3)),
+              {Inexact, Rounded})
+
+        # underflow to zero signals the full set of conditions; a trap
+        # on any of them must not suppress the flags of the others
+        check(Underflow,
+              lambda c: c.divide(Decimal('1E-999'), Decimal('1E+10')),
+              {Clamped, Inexact, Rounded, Subnormal, Underflow})
+        check(Subnormal,
+              lambda c: c.divide(Decimal('1E-999'), Decimal('1E+10')),
+              {Clamped, Inexact, Rounded, Subnormal, Underflow})
+
+        # divmod by zero signals DivisionByZero (for the quotient) and
+        # InvalidOperation (for the remainder)
+        check(DivisionByZero,
+              lambda c: c.divmod(Decimal(5), Decimal(0)),
+              {DivisionByZero, InvalidOperation})
+
+        # to_integral_exact signals Inexact and Rounded together
+        check(Inexact,
+              lambda c: c.to_integral_exact(Decimal('2.5')),
+              {Inexact, Rounded})
+
     def test_flag_comparisons(self):
         Context = self.decimal.Context
         Inexact = self.decimal.Inexact
@@ -4234,6 +4287,56 @@ def test_signaldict_repr(self):
 class PyContextFlags(ContextFlags, unittest.TestCase):
     decimal = P
 
+    def test_deferred_traps_attributes(self):
+        # The _defer_traps state lives in the _deferred_traps and
+        # _deferred_error attributes, which __setattr__ must accept.
+        Context = self.decimal.Context
+        Decimal = self.decimal.Decimal
+        Overflow = self.decimal.Overflow
+        context = Context(prec=9, Emin=-999, Emax=999, traps=[])
+        self.assertEqual(context._deferred_traps, 0)
+        self.assertIsNone(context._deferred_error)
+        context._deferred_traps = 1
+        context._deferred_traps = 0
+        context._deferred_error = Overflow('test')
+        context._deferred_error = None
+        with self.assertRaises(AttributeError):
+            context._no_such_attribute = 1
+
+        # A trapped operation resets the state, so the context
+        # remains usable afterwards.
+        context.traps[Overflow] = 1
+        with self.assertRaises(Overflow):
+            context.multiply(Decimal('1E+800'), Decimal('1E+800'))
+        self.assertEqual(context._deferred_traps, 0)
+        self.assertIsNone(context._deferred_error)
+
+    def test_deferred_traps_exception_discards_error(self):
+        # If an exception propagates through a _defer_traps block, the
+        # deferred error is discarded at every nesting level and the
+        # state is reset, so the propagating exception wins and the
+        # context stays usable.
+        Context = self.decimal.Context
+        Decimal = self.decimal.Decimal
+        Overflow = self.decimal.Overflow
+        context = Context(prec=9, Emin=-999, Emax=999, traps=[Overflow])
+        with self.assertRaises(RuntimeError):
+            with context._defer_traps():
+                with context._defer_traps():
+                    context._raise_error(Overflow, 'test', 1)
+                    raise RuntimeError('boom')
+        self.assertEqual(context._deferred_traps, 0)
+        self.assertIsNone(context._deferred_error)
+
+        # The flag was still set, and a later trapped operation
+        # raises its own error rather than the discarded one.
+        self.assertTrue(context.flags[Overflow])
+        context.clear_flags()
+        with self.assertRaises(Overflow):
+            context.multiply(Decimal('1E+800'), Decimal('1E+800'))
+        self.assertEqual(context._deferred_traps, 0)
+        self.assertIsNone(context._deferred_error)
+
 class SpecialContexts:
     """Test the context templates."""
 
diff --git a/Misc/NEWS.d/next/Library/2026-10-30-09-39-28.gh-issue-160945.b3kR7x.rst b/Misc/NEWS.d/next/Library/2026-10-30-09-39-28.gh-issue-160945.b3kR7x.rst
new file mode 100644
index 00000000000000..ca327effa7ba8c
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2026-10-30-09-39-28.gh-issue-160945.b3kR7x.rst
@@ -0,0 +1,5 @@
+Fixed :mod:`!_pydecimal` losing the context flags of companion conditions
+when a signal was trapped. After an operation that signals several
+conditions at once (for example an overflowing multiplication), the flags
+for *all* encountered conditions are now set, matching :mod:`!_decimal`;
+the exception raised is unchanged.