sb-kernel::%type-union stack exhaustion on bad defun

The following code does not drop me into the debugger in sbcl2.3.6

(defun check(expr &key predicate)
  (if predicate
      (list (check expr) predicate)
      (if (atom expr) t
   (check (first expr)))))

(defmacro ucheck(expr predicate)
  (check expr :predicate predicate))

(defconstant +comparisonops+ `(,#'< = ,#'= ,#'eql ,#'equalp))

(defun es() (ucheck 5 (member w +comparisonops+)))

but in versions 2.3.7 through 2.3.9 I get:

 CHECK
* UCHECK
* +COMPARISONOPS+
* INFO: Control stack guard page unprotected
Control stack guard page temporarily disabled: proceed with caution

debugger invoked on a SB-KERNEL::CONTROL-STACK-EXHAUSTED in thread
#<THREAD tid=2155644 "main thread" RUNNING {10044A00A3}>:
  Control stack exhausted (no more space for function call frames).
This is probably due to heavily nested or infinitely recursive function
calls, or a tail call that SBCL cannot or has not optimized away.

PROCEED WITH CAUTION.

Type HELP for debugger help, or (SB-EXT:EXIT) to exit from SBCL.

restarts (invokable by number or by possibly-abbreviated name):
  0: [ABORT] Exit debugger, returning to top level.

(SB-KERNEL::CONTROL-STACK-EXHAUSTED-ERROR)
0]

Looking through the backtrace I see a lot of:
1079: (SB-KERNEL:TYPE-UNION #<SB-KERNEL:CONS-TYPE (CONS (AND (NOT (FUNCTION (NUMBER T) (VALUES BOOLEAN . #1=#))) (FUNCTION (T T) (VALUES BOOLEAN . #1#)) (NOT (FUNCTION (NUMBER &REST T) (VALUES BOOLEAN . #1#)))) LIST)> #<SB-KERNEL:CONS-TYPE (CONS (FUNCTION (NUMBER T) (VALUES BOOLEAN &OPTIONAL)) LIST)> #<SB-KERNEL:CONS-TYPE (CONS (OR (MEMBER =) (AND (NOT (FUNCTION # #)) FUNCTION)) CONS)>) [more]
1080: (SB-KERNEL::%TYPE-UNION2 #<SB-KERNEL:CONS-TYPE (CONS (OR (FUNCTION (NUMBER T) (VALUES BOOLEAN . #1=(&OPTIONAL))) (AND (NOT (FUNCTION # #)) (FUNCTION (T T) (VALUES BOOLEAN . #1#)) (NOT (FUNCTION # #)))) LIST)> #<SB-KERNEL:CONS-TYPE (CONS (OR (MEMBER =) (AND (NOT (FUNCTION # #)) FUNCTION)) CONS)>)
1081: (SB-KERNEL::TYPE-UNION2 #<SB-KERNEL:CONS-TYPE (CONS (OR (FUNCTION (NUMBER T) (VALUES BOOLEAN . #1=(&OPTIONAL))) (AND (NOT (FUNCTION # #)) (FUNCTION (T T) (VALUES BOOLEAN . #1#)) (NOT (FUNCTION # #)))) LIST)> #<SB-KERNEL:CONS-TYPE (CONS (OR (MEMBER =) (AND (NOT (FUNCTION # #)) FUNCTION)) CONS)>)
1082: (SB-KERNEL::SIMPLIFY-UNIONS (#<SB-KERNEL:CONS-TYPE (CONS (OR (FUNCTION (NUMBER T) (VALUES BOOLEAN &OPTIONAL)) (AND (NOT #) (FUNCTION # #) (NOT #))) LIST)> #<SB-KERNEL:CONS-TYPE (CONS (OR (MEMBER =) (AND (NOT #) FUNCTION)) CONS)>))
1083: (SB-KERNEL::%TYPE-UNION (#<SB-KERNEL:CONS-TYPE (CONS (AND (NOT (FUNCTION # #)) (FUNCTION (T T) (VALUES BOOLEAN . #1=#)) (NOT (FUNCTION # #))) LIST)> #<SB-KERNEL:CONS-TYPE (CONS (FUNCTION (NUMBER T) (VALUES BOOLEAN . #1#)) LIST)> #<SB-KERNEL:CONS-TYPE (CONS (OR (MEMBER =) (AND (NOT #) FUNCTION)) CONS)>))

uname -a gives:
Linux papias 6.2.0-33-generic #33~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu Sep 7 10:33:52 UTC 2 x86_64 x86_64 x86_64 GNU/Linux

I understand the above code is clearly buggy, but I don't think sbcl should exhaust the stack in this case.