lock.cpp\post_blockage() could enter endless loop if, while delivering in-process AST, next blocking request will be freed by its owner.
It was reproduced when testing nbackup. I see no chance to this bug to occur in ClassicServer because it not released LM's mutex
in signal_owner() rountine so blocking requests queue is stable.
While i didn't tested v2.0.x i believe its affected too.
v2.5 is not affected as it already have similar fix made during refactoring LM internals.