The aim of this article is to show vacillation in using the variants of forms and derivatives of several rare and inflectionally irregular verbs with stems 'przasc' (spin) and 'bósc' (butt). It consists of three parts: two of them deal with the alternants of verbal stems in the passive participle (e.g. for 'chciec - chcian-y) and the impersonal finite form (chcian-o), the last with the imperative for verbs with stem 'bósc'. Two possibilities for participle alternants for these verbs are: old soft 'przedzion-y (bodzion-y)' and new hardened 'przedzon-y (bodzon-y)'. These two variants co-occur in both verbal stems for centuries. Because of their rarity users often cannot decide which one is correct or better. Polish historical dictionaries reflect this situation, also visible from the data found in linguistic corpora. In spite of this, Polish normative dictionaries and language handbooks assess categorically some participles and impersonal forms as either correct or incorrect. The inconsistency of such codificational decisions and advice is visible among the various sources and sometimes in the same book. In the last part of the article a similar sort of decision in the well-known Polish spelling dictionary published by the leading Polish scholarly publisher PWN is discussed. This dictionary gives two variants for the imperative of verbs with the stem 'bósc' and connects the distinction with grammatical number: 'bódz' but 'bodzcie'. There are no arguments for such codification because these forms are so rare that it is impossible to find any occurrence of them in the corpora (although they are semantically and structurally possible). Other dictionaries and handbooks repeat this advice, often inconsistently - only in some of the centric for such verbs. The author's view is that codification of standard Polish should be based on linguistic data, taken from corpora. In many cases it is much better not to decide which variant is the only acceptable one or the preferred one, because we lack foundations for any decision. This is especially true when the given phenomenon is very rare and the structure of the language provides for the existence of variants.