WRT performance, I specifically wrote "truly inadequate" in an attempt to distinguish between breaking and merely undesirably bad performance, but that is a somewhat vague line - unavoidably, I think.
Off the top of my head, the only sort of security 'problem' that can safely be put aside would be one that cannot lead to serious or irreversible harm, or requires the organization to have already been terminally compromised.
That I have to make these distinctions is, in fact, the basis of my concerns. They follow from two premises: firstly, the purpose of an inspection is to find problems that need to be fixed, and secondly, whether a problem falls into that category is orthogonal to the questions of what functional, structural or operational categories it falls in.
If your organization is such that you cannot even discuss these issues until you have working code, then you have other, bigger, problems.
Of all the issues that might be raised in a review, which are the most important to fix? I find it difficult to imagine a ranking in which incorrect functioning (including security vulnerabilities and truly inadequate performance) are not at the top (answering the ranking question with "all of them" would just be a way to avoid the issue.)
In this pyramid, issues of this nature are to be found at all levels except the top one, mixed in with more-or-less subjective ones, such as "Is a new API generally useful and not overly specific?" - pointless flame wars have erupted over issues such as this (also, orthogonally, code review is late in the game to be asking this one.)
For a review to be successful, its participants need to be able to restrain themselves from going down rabbit holes that could not lead to necessary rework - or a strong moderator, if the individual participants cannot act with adequate self-restraint.