While the reporter's problem is observed in SLIME, there is indeed an underlying issue in SBCL: when SB-DI:PREPROCESS-FOR-EVAL wraps symbols in SYMBOL-MACROLET, it probably should take care on disabling package locks (as it's not the user or his code who violates the lock, it's SBCL debugger itself; signalling package lock violation in this case is confusing and unexpected).
While the reporter's problem is observed in SLIME, there is indeed an underlying issue in SBCL: when SB-DI:PREPROCES S-FOR-EVAL wraps symbols in SYMBOL-MACROLET, it probably should take care on disabling package locks (as it's not the user or his code who violates the lock, it's SBCL debugger itself; signalling package lock violation in this case is confusing and unexpected).