On pg. 171 fn. 15 I note that hiemps is a spelling attested in the 5th century Mediceus manuscript of Vergil. I could also have noted that the spelling with p is attested epigraphically, e.g. in the so-called Menologium Rusticum Colotianum (CIL I2.1 p. 281 col. 4 line. 9, Rome) and on the Index Nundinarius olim Fulvii Ursini (CIL 6.32505). The grammarians opposed this spelling because p did not show up in the rest of the stem. Terentius Scaurus wrote (Keil 7.21):
similiter (to sumtus, demptus, and comtus) hiems carere p littera debet, quia in ceteris casibus nusquam p nec b propinqua eius respondet, sine quarum altera nusquam in Latinis ea nomina declinatur, quae in ψ Graeca voce efferuntur, ut “princeps” et “caelebs” quia principis et caelibis scribitur.
Marius Victorinus (Keil 6.21) and Papirian. apud Cassiod. (Keil 7.161) make similar comments. In fact spellings without p in line with the grammarians recommendations are attested for SVMSIT (CIL 4.1940. ARACVSA PRVDENTE[R]/ SVMSIT SIBI CASTA MVTHVNIVM, 4.2067 line 10), SVMTVS (CIL 3.14607) and DIREMSIT (CIL 9.5036).
For a discussion of the phonetics of epenthetic or "emergent" stops see Ohala 1995:162–3.
Ohala, John. 1995. A probable case of clicks influencing the sound patterns of some European languages. Phonetica 52:160–170.